Rapid recovery and altered neurochemical dependence of locomotor central pattern generation following lumbar neonatal spinal cord injury

Abstract

Key points: Spinal compression injury targeted to the neonatal upper lumbar spinal cord, the region of highest hindlimb locomotor rhythmogenicity, leads to an initial paralysis of the hindlimbs. Behavioural recovery is evident within a few days and approaches normal function within about 3 weeks. Fictive locomotion in the isolated injured spinal cord cannot be elicited by a neurochemical cocktail containing NMDA, dopamine and serotonin 1 day post-injury, but can 3 days post-injury as readily as in the uninjured spinal cord. Low frequency coordinated rhythmic activity can be elicited in the isolated uninjured spinal cord by NMDA + dopamine (without serotonin), but not in the isolated injured spinal cord. In both the injured and uninjured spinal cord, eliciting bona fide fictive locomotion requires the additional presence of serotonin.

Abstract: Following incomplete compression injury in the thoracic spinal cord of neonatal mice 1 day after birth (P1), we previously reported that virtually normal hindlimb locomotor function is recovered within about 3 weeks despite substantial permanent thoracic tissue loss. Here, we asked whether similar recovery occurs following lumbar injury that impacts more directly on the locomotor central pattern generator (CPG). As in thoracic injuries, lumbar injuries caused about 90% neuronal loss at the injury site and increased serotonergic innervation below the injury. Motor recovery was slower after lumbar than thoracic injury, but virtually normal function was attained by P25 in both cases. Locomotor CPG status was tested by eliciting fictive locomotion in isolated spinal cords using a widely used neurochemical cocktail (NMDA, dopamine, serotonin). No fictive locomotion could be elicited 1 day post-injury, but could within 3 days post-injury as readily as in age-matched uninjured control spinal cords. Burst patterning and coordination were largely similar in injured and control spinal cords but there were differences. Notably, in both groups there were two main locomotor frequencies, but injured spinal cords exhibited a shift towards the higher frequency. Injury also altered the neurochemical dependence of locomotor CPG output, such that injured spinal cords, unlike control spinal cords, were incapable of generating low frequency rhythmic coordinated activity in the presence of NMDA and dopamine alone. Thus, the neonatal spinal cord also exhibits remarkable functional recovery after lumbar injuries, but the neurochemical sensitivity of locomotor circuitry is modified in the process.

Keywords: adaptive plasticity; network re-organization; recovery; spinal cord injury; sprouting.

Figure 1. Schematic diagram summarizing the different experiments performed in this study

A spinal cord compression injury (SCC) was generated in 1‐day‐old (P1) mice at thoracic (T8–10, Th‐SCC) or lumbar (T13–L2, Lb‐SCC) levels. The behavioural locomotor tests, air‐stepping (P2, P5 and P9) and linear swimming (P15 and P25) were performed on injured and uninjured age‐matched control mice. Spinal cords from Lb‐SCC and control mice were dissected out at P2, P3 and P4 and electrophysiological recordings were made simultaneously from right and left L2 roots (RL2 and LL2, respectively) and right and left L5 roots (RL5 and LL5, respectively). The isolated spinal cords were then fixed and processed for immunohistochemistry and confocal microscopy. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 2. Slower behavioural recovery occurs after Lb‐SCC than after Th‐SCC

A, kinematic analysis of hindlimb movements at P2, P5 and P9 showing hindlimb movement amplitudes in Th‐SCC, Lb‐SCC and age‐matched uninjured control mice. B, paw trajectory in injured mice as a percentage of that in P9 control mice. C, variation of the paw‐ankle‐knee (p‐a‐k) angle. D, percentage of flexion as assessed within the range of 0–45 deg of the p‐a‐k angle. Arrows represent nil values. E, minimum and maximum values of the p‐a‐k angle. Error bars represent SD. ^* P < 0.05; ^*** P < 0.01; ^**** P < 0.005; ^***** P < 0.001; ^ø P > 0.05.

Figure 3. Properties of neurochemically (NMDA/DA/5‐HT) induced fictive locomotion in neonatal control mice

A, raw traces from ventral root recordings showing a long‐lasting episode of rhythmic locomotor activity with left–right (RL2–LL2/RL5–LL5) and flexor–extensor (RL2–RL5/LL2–LL5) alternation. A rectified and smoothed version (red line) of the raw trace was used to determine burst onset and cycle period. B, autocorrelograms corresponding to each ventral root shown in A. C, cross‐correlograms shown for left–right and flexor–extensor ventral root pairs. D, circular graph showing phase lag relationships for RL2 versus LL5 (in phase) and LL2 versus LL5 (out of phase). Vectors show the mean phase values and r values, their orientations indicate preferred phase of firing and their lengths are proportional to coupling strength. E, scatter diagrams of burst duration versus cycle period, with linear regression lines indicated.

Figure 4. Transient and reversible loss of fictive locomotion after Lb‐SCC

A, representative recordings from RL2 and LL2 ventral roots 1 day post‐injury immediately following application of the NMDA/DA/5‐HT cocktail used to trigger fictive locomotion. B, recordings from RL2 and LL2 ventral roots showing lack of coordinated rhythmic activity at P2 in Lb‐SCC preparations. C, higher temporal resolution sequence from the RL2 ventral root recording shown in B. Activity is not coordinated into rhythmic bursts. D, diagram showing the success rate for triggering fictive locomotion in Lb‐SCC and in age‐matched uninjured control spinal cords (P2, P3 and P4). Note that even control spinal cords do not always exhibit rhythmic activity in all 4 ventral roots.

Figure 5. Fictive locomotion has similar properties at P4 in Lb‐SCC and control preparations

A, representative ventral root recordings P4 from an injured spinal cord show obvious rhythmicity characteristic of fictive locomotion (FL). Rectified and smoothed versions of the raw traces are shown in red. B, comparison of rhythm strength for different ventral root pairs in control and Lb‐SCC preparations at P4. There was no statistically significant difference between rhythm strength of control and Lb‐SCC recordings for the RL2/LL5, LL2/LL5 and RL5/LL5 ventral root pairs (^ø P > 0.05). Error bars represent SD. C, circular plots showing the preferred phase relationships for different ventral root pairs at P4 (red line, injured; blue line, controls). Vector orientation and error bars represent mean and SD, respectively, of the phase for each indicated ventral root pair. D, representative ventral root recordings showing low and high bursting frequency in control (left panel) and Lb‐SCC (right panel) spinal cord preparations. E, distribution of rhythm frequencies in different roots during fictive locomotion at P4 (red line, injured; blue line, controls).

Figure 6. Neuronal loss and sprouting of serotonergic fibres at P4 in Lb‐SCC

Representative confocal images from transverse sections of uninjured control (A) and Lb‐SCC (B) spinal cords immunostained for NeuN (green) and counterstained with Hoechst 33258 (blue). C, quantification of the total number of NeuN‐positive cells in Lb‐SCC and age‐matched uninjured control spinal cord sections at the level of injury shown as percentage of control. D–J, confocal images from spinal cord cross‐sections from control (D and G) and Lb‐SCC (E, F, H, I and J) spinal cords immunostained for serotonin (5‐HT). G–J show higher magnification images from the regions indicated by white rectangles in D–F. K, quantification of the 5‐HT‐immunopositive pixel density in the control and Lb‐SCC sections over the entire section (whole) and in selected regions of interest (ROIs) containing the highest pixel density. Error bars represent SD. ^* P < 0.05; ^** P < 0.025; ^ø P > 0.05; DH, dorsal horn; VH, ventral horn.

Figure 7. Increased serotonin dependence of hindlimb locomotor CPG output following Lb‐SCC

A–D, serotonin dependence in control spinal cords. Representative raw ventral root recordings from control spinal cords prior to (A) and after serotonin application (C) during rhythmic activity elicited by a reduced neurochemical cocktail (NMDA, 5 μm; DA, 50 μm). B and D, auto‐ and cross‐correlations for the traces shown in A and C, respectively. A slow rhythmic output can be elicited without serotonin, but addition of serotonin increases frequency markedly to match that of bona fide fictive locomotion. E–H, serotonin dependence in P4 Lb‐SCC preparations. Representative raw ventral root recordings from a Lb‐SCC spinal cord prior to (E) and after serotonin application (G) while in the presence of the reduced neurochemical cocktail (NMDA, 5 μm; DA, 50 μm). F and H, auto‐ and cross‐correlations for the traces shown in E and G, respectively. The non‐oscillatory, decaying cross‐correlations in F indicate a lack of rhythmicity in the absence of serotonin. Rhythmicity is instated following serotonin application, as shown by the oscillatory auto‐ and cross‐correlation functions in H.

Figure 8. Dose dependence of NMDA on the rhythmic output of the locomotor CPG in isolated spinal cords from P4 Lb‐SCC and uninjured control mice

Representative LL2 root activity elicited by exposure to a constant DA concentration (50 μm) with step increases in NMDA concentration (5–8 μm), with 10 μm serotonin added as a last step, in P4 control (A) and Lb‐SCC (B) preparations. Autocorrelation of rhythmic activity in RL2, LL5 and LL2 ventral roots with step increase in NMDA concentration and subsequent addition of serotonin in P4 control (C) and P4 Lb‐SCC (D) preparations. E, NMDA dose–response on cross‐correlation coefficient (CCC) integrated by time for the ventral root pair LL2/LL5. F, occurrence of coordinated rhythmic activity in control (n = 8) and Lb‐SCC (n = 9) preparations with step increases of NMDA concentration and subsequent addition of serotonin. The data represent the percentage of coordinated activity within each group. Red arrows indicate nil value.

Figure 9. Nearly complete functional recovery of linear swimming within 3 weeks post‐injury by Th‐SCC and Lb‐SCC mice

A, representative hindpaw trajectories and kinematics of the paw, ankle, knee, hip and iliac crest during linear swimming at P15 and P25 in injured and control mice. B, mean hindpaw trajectory expressed as percentage of age‐matched control mice. C, instantaneous variation of paw velocity. D, paw trajectory as percentage of age‐matched uninjured control mice. E, instantaneous variation of paw‐ankle‐knee angle. F, minimum and maximum paw‐ankle‐knee angles. G, number of limb strokes. Error bars represent SD. ^** P < 0.025; ^*** P < 0.01; ^ø P > 0.05.