Article Muscle Spindle Feedback Directs Locomotor Recovery and Circuit Reorganization after Spinal Cord Injury Aya Takeoka,1,2,4 Isabel Vollenweider,3,4 Gre´goire Courtine,3,5 and Silvia Arber1,2,5,* 1Biozentrum,
Department of Cell Biology, University of Basel, 4056 Basel, Switzerland Miescher Institute for Biomedical Research, 4058 Basel, Switzerland 3Brain Mind Institute and Centre for Neuroprosthetics, Ecole Polytechnique Fe ´ de´rale de Lausanne (EPFL), 1015 Lausanne, Switzerland 4Co-first author 5Co-senior author *Correspondence:
[email protected] http://dx.doi.org/10.1016/j.cell.2014.11.019 2Friedrich
SUMMARY
Spinal cord injuries alter motor function by disconnecting neural circuits above and below the lesion, rendering sensory inputs a primary source of direct external drive to neuronal networks caudal to the injury. Here, we studied mice lacking functional muscle spindle feedback to determine the role of this sensory channel in gait control and locomotor recovery after spinal cord injury. High-resolution kinematic analysis of intact mutant mice revealed proficient execution in basic locomotor tasks but poor performance in a precision task. After injury, wildtype mice spontaneously recovered basic locomotor function, whereas mice with deficient muscle spindle feedback failed to regain control over the hindlimb on the lesioned side. Virus-mediated tracing demonstrated that mutant mice exhibit defective rearrangements of descending circuits projecting to deprived spinal segments during recovery. Our findings reveal an essential role for muscle spindle feedback in directing basic locomotor recovery and facilitating circuit reorganization after spinal cord injury. INTRODUCTION Spinal cord injury has an immediate and devastating impact on the control of movement. The origin of motor impairments lies in the physical disconnection of descending pathways from spinal circuits below the lesion, depriving them of synaptic input essential for the generation and regulation of motor output. Despite the failure of severed axons to regenerate at long distance (Ramon y Cajal, 1928; Tello, 1907), partial lesions of the human spinal cord are frequently associated with spontaneous functional improvement (Curt et al., 2008). One of many challenges in restoring motor control after spinal cord injury is to re-establish a sufficient level of task-specific excitability within disconnected local spinal circuits to drive motor neurons caudal to the injury. 1626 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
Recent studies on incomplete spinal cord injury animal models uncovered some of the mechanisms that may contribute to spontaneous motor recovery (Ballermann and Fouad, 2006; Bareyre et al., 2004; Courtine et al., 2008; Jankowska and Edgley, 2006; Rosenzweig et al., 2010; Zo¨rner et al., 2014). These investigations showed that recovery correlates with the establishment of intraspinal detour circuits. Such alternative pathways through the spared tissue form novel functional bridges across the lesioned spinal segments. At present, circuit-level mechanisms promoting the formation of detour circuits to restore control of movement remain elusive, even though such insight might play a pivotal role in developing interventions that enhance locomotor recovery after spinal cord injury. Various studies suggest that sensory information plays a critical role in gait control and in locomotor recovery after spinal cord injury (Edgerton et al., 2008; Pearson, 2008; Rossignol et al., 2006; Rossignol and Frigon, 2011; Windhorst, 2007). The most common medical practice used to facilitate motor recovery of paraplegic patients is weight-supported locomotor rehabilitation (Dietz and Fouad, 2014; Knikou and Mummidisetty, 2014; Roy et al., 2012). Repetitive movement during rehabilitative training likely enhances glutamatergic dorsal root ganglia (DRG) sensory feedback, which constitutes the primary extrinsic source of excitation entering the spinal cord below injury to engage local spinal circuits. This interpretation is supported by evidence from animal models in which spinal cord injury coupled to partial or complete elimination of sensory input impairs gait control and locomotor recovery (Bouyer and Rossignol, 2003; Lavrov et al., 2008). However, the DRG neuron subtype promoting locomotor recovery and the mechanisms by which this process takes place are unclear. Proprioceptive sensory neurons innervate sense organs in the muscle and transmit information about muscle contraction to the spinal cord (Brown, 1981; Rossignol et al., 2006; Windhorst, 2007). Their influence on the activity of central circuits is essential for modulation and adjustment of motor output (Pearson, 2008; Rossignol et al., 2006). Muscle spindle afferents constitute a subset of proprioceptors contacting muscle spindle sense organs. They exhibit the most widespread central projection pattern of all DRG sensory neurons and establish
synaptic contacts with motor neurons and various classes of interneurons implicated in motor control (Brown, 1981; Eccles et al., 1957; Rossignol et al., 2006; Windhorst, 2007). Muscle spindle afferents are thus in a prime position to convey direct excitation to spinal circuits relevant to the regulation of motor behavior, especially under conditions of disconnected descending input. The zinc-finger transcription factor Egr3 is expressed selectively by muscle spindle-intrinsic intrafusal muscle fibers, and its mutation results in early postnatal degeneration of muscle spindles (Tourtellotte and Milbrandt, 1998). This defect abolishes normal function of muscle spindle afferents as assessed electrophysiologically (Chen et al., 2002) and leads to gait ataxia (Tourtellotte and Milbrandt, 1998). Egr3 mutant mice thus represent a genetic model with DRG sensory neuron dysfunction selectively restricted to muscle spindle afferents. They provide an opportunity to investigate how this feedback channel contributes to gait control in intact mice and influences locomotor recovery and circuit reorganization after spinal cord injury. To address this question, we conducted kinematic analyses in wild-type and Egr3 mutant mice. Deficiency of muscle spindle feedback did not affect basic motor abilities in intact Egr3 mutant mice beyond specific gait features. However, lack of muscle spindle feedback severely restricted spontaneous recovery after incomplete spinal cord injury. Egr3 mutant mice also exhibit a markedly reduced ability to establish descending detour circuits restoring access to spinal circuits below spinal cord injury. We conclude that muscle spindle feedback is a key neuronal substrate to direct circuit rearrangement necessary for locomotor recovery after incomplete spinal cord injury. RESULTS Proficient Basic Locomotion in Absence of Muscle Spindle Feedback We performed high-resolution video recordings to reconstruct hindlimb kinematics in wild-type and Egr3 mutant mice (Figures 1A and 1B). We focused on task-dependent contributions of muscle spindle input to hindlimb motor control with the aim to establish a baseline to which we could compare the locomotor recovery process after spinal cord injury. We first assessed hindlimb motor control during basic overground locomotion. Wild-type and Egr3 mutant mice performed this task with reciprocal activation of flexor and extensor muscles and alternation between left and right hindlimbs (Figure 1B; Movie S1 available online). However, Egr3 mutant mice exhibited gait ataxia as reported previously (Tourtellotte and Milbrandt, 1998). To characterize gait patterns, we computed >100 kinematic parameters that provide a comprehensive quantification of locomotor features (Figure S1) (Courtine et al., 2008). We subjected all measured parameters to a principal component (PC) analysis (van den Brand et al., 2012) (averaged values of 10–25 step cycles/hindlimb/mouse; n = 22 wild-type and n = 19 Egr3 mutants; Figure S2). We then visualized gait patterns in the new space created by PC1–3, where PC1 explained the highest variance (18%) and distinguished the two genotypes (Figure 1C).
The locomotor phenotype observed in Egr3 mutant mice was limited to distinct gait features represented in PC1 and approximately 65% of all parameters did not correlate with this genotype-specific PC1 (Figure S3A). To evaluate the ability of Egr3 mutant mice to adjust gait patterns to changing locomotor velocities, we tested mice during stepping on a treadmill. Both wild-type and Egr3 mutant mice were capable of stepping across the entire range of tested treadmill speeds (7–23 cm/s; Figures 1D and S3B). PC1 captured adjustment of gait patterns with increasing speed in mice of both genotypes (16% of explained variance; Figures 1D and S3C), whereas PC2 segregated genotypic differences independent of velocity (10% explained variance; Figures 1D and S3C). Electromyogram (EMG) recordings of ankle extensor and flexor muscles revealed that both genotypes showed appropriate speed-dependent adjustments in burst duration (Figure 1D). These findings resonate with work demonstrating that the flexion phase of the step cycle remains constant, whereas the extension phase progressively shortens with increased locomotor speed (Arshavskiĭ et al., 1965; Halbertsma, 1983), a property we now demonstrate to be independent of muscle spindle sensory feedback. In summary, both wild-type and Egr3 mutant mice are able to perform basic locomotor tasks proficiently, but mutant mice display specific gait alterations concordant with the previously proposed role of muscle spindle feedback in control and adjustment of locomotion (Pearson, 2008; Rossignol et al., 2006; Windhorst, 2007). Muscle Spindle Feedback Is Essential for Locomotor Precision Task and Swimming Next, we tested mice of both genotypes during walking on a horizontal ladder, requiring precision in foot placement. Whereas wild-type mice progressed across the ladder with ease, Egr3 mutants frequently slipped off or missed rungs, which was reflected in aberrant bouts of EMG activity (Figure 2A; Movie S2). Quantification of foot positioning relative to successive rungs revealed that wild-type mice targeted rungs precisely, whereas Egr3 mutant mice displayed near-random foot placement (Figure 2B). These findings demonstrate an essential role for muscle spindle feedback circuits in the regulation of accurate foot placement in a locomotor precision task. Egr3 mutant mice exhibit selective defects of muscle spindle feedback, but other sensory feedback is preserved (Tourtellotte and Milbrandt, 1998). During swimming, afferents from Golgi tendon organs are attenuated due to reduced weight load (Gruner and Altman, 1980). Proprioceptive signaling therefore relies almost exclusively on muscle spindle feedback. We found that during swimming, wild-type mice displayed well-coordinated alternation of left and right hindlimbs with reciprocal activity of ankle flexor and extensor muscles (Figures 2C and 2D). In contrast, Egr3 mutant mice were unable to keep afloat and showed uncoordinated hindlimb movements with extensive coactivation of antagonistic muscles (Figures 2C and 2D). Together, these findings stress the pivotal function of muscle spindle feedback in the control of swimming, a condition when Golgi tendon organ and cutaneous feedback circuits only play a limited task-related function. Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. 1627
Egr3-/-
A Wild-type
B
DRG
1 cm
Muscle MN
5 cm Extensor Flexor Left stance Right stance Time (s) 0
MN
Motor neurons Muscle spindle afferents Intrafusal muscle fiber
Runway
1 mV 1 mV 0.5
2%
**
(1 )
***
2
-10 -10
0 1( PC
) 18%
2
Burst duration (s)
15
0
0
0.5
1
Slopes of regression lines
Speed - dependent muscle recruitment
0
-10 10
1
Treadmill PC1 (16%): Speed (a.u.) 8 4 0 -4 -8
15
PC
PC1: Genotype (a.u.) 4
D
*
PC3 (10%)
C
Egr3-/-
Wild-type Iliac crest Stance Swing Hip Endpoint Knee trajectory Ankle Foot
1.5
1
n.s.
Extensor
1
0.5
0.5 Flexor
0 0
n.s.
0
0.5 1 1.5 Cycle duration (s)
-0.5
Extensor Flexor
PC3
0 -2 -4
PC2 (10%): Genotype (a.u.)
***
2
Egr3-/-
3 0
PC
Wild-type
6
PC1
-3
Wildtype 7 11 15 19 23 cm/s Egr3-/7 11 15 19 23 cm/s
**
Figure 1. Proficient Basic Locomotion in Absence of Muscle Spindle Feedback (A) Egr3 mutation results in selective degeneration of muscle spindles and nonfunctional muscle spindle feedback circuits. (B) Color-coded stick decomposition of hindlimb movement during three consecutive steps with limb endpoint trajectories and velocity vector at swing onset during basic overground locomotion in both genotypes (EMG activity of an extensor and a flexor muscle displayed below; dark gray bars, stance; empty spaces, swing). (C) PC analysis was applied on 103 gait parameters measured during overground locomotion (10–25 gait cycles/hindlimb/mouse, n = 22 wild-type and n = 19 Egr3 mutants). Gait cycles are represented for each animal and hindlimb (individual dots) in the new space created by PC1–3. Least-squares elliptical fitting (95% confidence) was computed to emphasize differences between genotypes. Histogram plot, mean values of PC1 scores for each genotype. (D) PC analysis applied on averaged values of 108 gait parameters (15–30 gait cycles/mouse/speed, n = 10 for each genotype) measured during stepping on a treadmill at five different speeds (7–23 cm/s). Histogram plots, mean values of PC1 and PC2 scores. Correlation between step-cycle duration and extensor or flexor burst duration. Each regression line was computed separately for a given animal (n = 4 for each genotype; 25–30 step cycles/mouse). Histogram plots, slopes of regression lines for extensor and flexor muscles. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant; error bars, SEM; extensor, gastrocnemius medialis; flexor, tibialis anterior; a.u., arbitrary unit. See also Figures S1, S2, and S3 and Movie S1.
Muscle Spindle Feedback Circuits Are Essential for Locomotor Recovery after Injury The core ability to perform basic locomotion is not disturbed in Egr3 mutant mice, providing an opportunity to assess the role of muscle spindle feedback circuits in gait control and spontaneous recovery after spinal cord injury. We placed a lateral hemisection injury at the thoracic level (T10) and confirmed lesion completeness upon termination of experiments (Figure 3A). This lesion interrupts descending tracts projecting ipsilateral to 1628 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
lesion (ipsilesional hereafter), which normally innervate lumbar segments containing circuits essential for the control of ipsilesional hindlimb muscles (Figure 3A). We performed kinematic analysis at regular intervals after injury to follow locomotor recovery (Figures 3A and 3B). Three days after injury (acute), both wild-type and Egr3 mutant mice dragged the ipsilesional hindlimb along the runway as they moved forward (Figure 3B; Movies S3 and S4). Wild-type mice gradually regained locomotor proficiency over the time course
Figure 2. Muscle Spindle Feedback Is Essential for Precision Tasks and Swimming
Horizontal ladder A
Egr3-/-
Wild-type
Hit Slip Miss 2 cm
Swing
2 cm 100 deg
Hindlimb Extensor
1 mV
Flexor
2 mV
Stance
Percentage of steps
B
25
rung
Miss 1 ±0.3%
Slip 2 ±1%
30 ±4%
37 ±2% 25
20
20
15
Hit 97 ±1%
10 5
0 -50 -25 0 25 50 Horizontal paw placement (%)
d D Paw placement = (d/D)*100
15
33 ±2%
10 5 0 -50 -25 0 25 50 Horizontal paw placement (%)
Swimming Limb trajetory
2 cm
Flexor Extensor 100 deg
Hindlimb R stroke L stroke Time (s)
0
0.5
0
Reciprocal Left - right activation alternation out-of phase
0 Flexor 1
0.5
CoLoss of activation alternation 1
out-of phase
Extensor
low
1
Extensor
Density
high
in-phase
0 Flexor 1
100 75 50 25 0 ***
Forward propulsion per stroke (cm)
2 cm
Endpoint trajetory
Power stroke Return stroke
Co-contraction of antagonists (%)
**
D
Egr3-/-
Wild-type
Left-right coordination (a.u.)
C
(A) Stick diagram decomposition of hindlimb movement for a representative wild-type and Egr3 mutant mouse during crossing of an elevated horizontal ladder with rungs (spacing 2 cm; below: hindlimb oscillation and traces of ankle extensor and flexor muscles for same mouse; dark gray bars, stance). (B) Bar graph quantifying relative positioning of hindpaws with respect to rung positions. Pie charts summarize total percentage of hits, slips, and misses (n = 9 mice per genotype; 259 steps for wild-type and 323 steps for Egr3 mutant mice). (C) Stick decomposition of hindlimb movement for a wild-type and Egr3 mutant mouse during swimming (below: limb endpoint trajectories, limb endpoint velocity vectors at power stroke onset, and raw traces of muscle activity for an extensor and flexor muscle together with hindlimb oscillations; dark gray bars, return stroke). Density plot displays coordination between antagonistic muscles during the represented trial (L-shaped patterns, reciprocal muscle activation; diagonal: continuous coactivation). Polar plot, coordination between left and right hindlimb oscillations (black lines, single gait cycle; red arrow, average of all gait cycles). (D) Histogram plots report mean values for representative kinematic and muscle activity-related variables extracted from PC analysis (n = 181 swim strokes, 10–15 strokes/mouse, n = 8 wild-type and n = 7 Egr3 mutant mice) during swimming task. *p < 0.05; **p < 0.01; ***p < 0.001; error bars, SEM; extensor, vastus lateralis; flexor, tibialis anterior; a.u., arbitrary unit. See also Figures S1 and S2 and Movie S2.
5 4 3 2 1 0
0 -0.2 -0.4 *
out-of -0.6 phase
Wild-type Egr3-/in-phase
Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. 1629
B
Lateral hemisection model
Wild-type
*
*
Dorsal and ventral view
3 days
Intact Iliac crest Stance Swing
1 cm
A
7 weeks
Acute
Chronic
Drag
Foot
T10
1 mV
5 cm Extensor Flexor Ipsi stance Contra stance
Egr3 -/-
Wild-type
15
PC3
PC3
12
-15 -10 PC1: Gradual re
15 10
Ipsi-lesional Hit: 1 ±1%
Miss: 87 ±4%
covery
Slip: 12 ±3%
% of steps
E
D
-15 2 PC
25 20 15 10 5 0
-12 -15
PC1: Little reco
very
rung
-50 -25 0 25 50 Horizontal paw placement (%)
10 10
P
-10 C2
*** ***
8
*** n.s.
4 0 -4 -8 -12
In ta Ac ct u 1 te w e 2 ek w e 4 ek w s C eek hr s on ic
C
3 days 1 wk 2 wks 4 wks 7 wks
Extensor Flexor Ipsi stance Contra stance
Wild-type Egr3-/-
Contra-lesional 18 ±5% 82 ±4%
% of steps
Intact
Chronic
Intermediate testings
1 mV
Baseline Hx Acute T10
PC1 (30%): Recovery (a.u.)
Glial scar
Egr3 -/-
CC
25 20 15 10 5 0
rung
-50 -25 0 25 50 Horizontal paw placement (%)
Figure 3. Muscle Spindle Feedback Is Essential to Direct Spontaneous Locomotor Recovery after Lateral Hemisection (A) Illustration of thoracic lateral hemisection model, including a representative lesion from a dorsal, ventral, and coronal view, and time line of experiment procedures. (B) For each genotype, a representative stick decomposition of hindlimb movement during basic overground locomotion is shown for intact, acute, and chronic time points for the same mouse (below: concurrent limb endpoint trajectory and velocity vector at swing onset, activity of an extensor muscle, and activity of a flexor muscle; dark gray bars, stance; red bars, dragging). (C) Representation of gait clusters in PC space for one mouse per genotype during intact stepping and at five different time points postinjury (10–15 steps per time point; 103 parameters per gait cycle). (D) Histogram plots reporting mean values of PC1 scores measured on all data combined (average of 10–25 steps per time point, 103 parameters per gait cycle, n = 9 wild-type mice, n = 7 Egr3 mutant mice). (E) Bar graph of relative positioning of hindpaws with respect to rung positions for chronically injured wild-type mice (n = 10). Pie charts summarize total percentage of hits, slips, and misses (n = 259 steps contralesional hindlimb; n = 147 steps ipsilesional hindlimb). *p < 0.05; ***p < 0.001; ns, not significant; error bars, SEM; extensor, medial gastrocnemius; flexor, tibialis anterior; Hx, hemisection; a.u., arbitrary unit; acute, 3 days postinjury; chronic, 7 weeks postinjury. See also Figures S1, S2, and S4 and Movies S3, S4, and S5.
1630 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
analyzed. By 7 weeks postinjury (chronic), they regained weightbearing plantar steps with regular alternation of stance and swing phases of the ipsilesional hindlimb (Figures 3B and S4A; Movies S3 and S4). In contrast, Egr3 mutant mice still exhibited severe locomotor deficits at the chronic stage (Figures 3B and S4A; Movies S3 and S4). To quantitatively assess the recovery of ipsilesional hindlimb function, we conducted a PC analysis comparing intact condition to each time point evaluated. We found that PC1 characterized the recovery process (30% explained variance; Figures 3C and 3D). In wild-type mice, time-dependent gait clusters gradually moved toward intact conditions, reflecting the progressive recovery of locomotor function (Figures 3C and 3D). In contrast, gait clusters of Egr3 mutant mice remained confined in the same PC space through the entire time course evaluated (Figures 3C and 3D). Detailed kinematic analysis revealed that in chronic wild-type mice, 73% of all parameters affected at acute stages improved significantly (p < 0.05) and 17% even recovered to levels measured before lesion (Figure S4A). Lack of locomotor recovery in Egr3 mutant mice was associated with persistent dragging of the ipsilesional hindlimb at all time points (Figures 3B and S4A). In addition, analysis of contralesional hindlimb gait parameters revealed that both wild-type and Egr3 mutant mice adjust their gait to ipsilesional hindlimb deficiencies similarly and immediately after lesion (Figure S4B). Together, our results demonstrate that defective muscle spindle feedback circuitry severely limits spontaneous locomotor recovery after incomplete spinal cord injury. Speed Adjustment, but Not Precision Control, Improves in Wild-Type Mice after Injury Next, we determined the extent to which hemisected wildtype mice regain the capacity to accommodate hindlimb movement to increasing walking speeds and to perform muscle spindle feedback-dependent swimming and ladder precision tasks. Wild-type mice at chronic stages recovered the ability to walk at the highest speed tested (23 cm/s). After a lack of ipsilesional muscle recruitment at acute stages, the modulation of ankle extensor muscle activity gradually recovered toward intact levels (Figure S4C). In contrast, prolonged paw dragging (Figure S4A) led to increased EMG bursts in ankle flexor muscles after lesion, a feature that only partially recovered at chronic stages (Figure S4C). Wild-type mice regained well-coordinated limb alternation during swimming (Figure S4D) (Zo¨rner et al., 2010), providing further evidence for recovery of basic locomotor features. During precision walking on the horizontal ladder at chronic stages, 87% of ipsilesional hindlimb steps resulted in a complete miss of the targeted rung, and 12% slipped off the rung. In contrast, most steps of the contralesional hindlimb were placed correctly on the rungs (82%) (Figure 3E; Movies S5). Taken together, these results indicate that after lateral hemisection injury, wild-type mice regain basic locomotor function but only partially recover speed-dependent adaptation and completely fail to recover precise paw placement required for ladder locomotion.
Muscle Spindle-Specific Feedback Needed for Functional Recovery Contrary to wild-type mice that regained the ability to move their ipsilesional hindlimb after injury, Egr3 mutants exhibited persistent lack of locomotor control. Because activity-dependent mechanisms contribute to recovery of locomotor function after spinal cord injury (Dietz and Fouad, 2014; Edgerton et al., 2008; Maier and Schwab, 2006), we next measured the degree of spontaneous motility in Egr3 mutant mice. We monitored home cage activity before injury and at regular intervals after lesion (Figure 4A). Both groups displayed decreased locomotor activity immediately after injury, but there were no significant genotype-related differences in distance covered throughout the recovery process (Figure 4A). We then asked whether daily administration of monoaminergic receptor agonists known to acutely enhance locomotor output in rodents with severe spinal cord injury (van den Brand et al., 2012) influence the recovery process in Egr3 mutant mice. We reasoned that despite indistinguishable motility between the two groups after lesion, spinal circuits in Egr3 mutants may be recruited less efficiently in the absence of functional muscle spindle feedback than in wild-type mice. We found that upon daily agonist administration, Egr3 mutants still exhibited an overall impediment in locomotor recovery (Figures 4B and 4C). The contribution of individual parameters to the recovery-associated PC1 showed a high correlation between spontaneous and daily drug administered groups for both genotypes (Figure 4D). These results demonstrate that muscle spindle sensory feedback is absolutely essential for directing the process of locomotor recovery after spinal cord injury and cannot be substituted for by daily activation of spinal circuits through pharmacological means. Muscle Spindle Feedback Promotes Efficient Detour Circuit Establishment around Lesion Because reorganization of supraspinal and intraspinal descending circuits parallels spontaneous recovery after incomplete spinal cord injury (Bareyre et al., 2004; Courtine et al., 2008; Rosenzweig et al., 2010; Zo¨rner et al., 2014), we asked whether presence of muscle spindle feedback influences these injuryinduced circuit reorganization responses. The formation of functional detour circuits relies on the ability of neuronal subpopulations to establish new connections to ipsilesional spinal circuits below the lesion. A predicted hallmark of such neurons is that they must have projections to segments below injury prior to lesion and establish novel synaptic connections after injury. To identify sources of such neurons, we performed a mapping approach to label neurons with projections to the ipsilesional lumbar spinal cord, by injection of G protein-deficient rabies viruses encoding fluorescent marker proteins (FP) (Rab-FP; Figure 5A) (Wickersham et al., 2007). We analyzed the relative abundance and pattern of marked neurons above lesion in intact, acute, and chronic mice. We first visualized descending supraspinal projection neurons in intact wild-type and Egr3 mutant mice. Retrogradely labeled neuron distribution was reminiscent of patterns of mapped premotor brainstem nuclei (Esposito et al., 2014), with most RabFPON neurons located in the magnocellular, followed by pontine, Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. 1631
Egr3-/-
Wild-type
A Intact
Acute
Intact
Chronic
Acute
150
Chronic
Distance covered during 20 min home cage monitoring (m)
30 cm
75
Wild-type Egr3-/-
In ta ct 4 da 1 y w s ee k 2 w ee ks 3 w ee ks 4 w ee ks 5 w ee ks 6 w ee ks 7 w ee ks
0
15 cm
7
Intact Acute 1 week 2 weeks 4 weeks Chronic
0 -5 n.s.
-10
2 cm
D
Wild-type
Correlation plots between PC1 factor loadings 1 r = 0.90
Spontaneous
Foot 1 cm
7
40°
Intact Acute 1 week 2 weeks 4 weeks Chronic
0 -5
-10
0
-1 -1 -0.5 0 0.5 1 Daily agonist exposure 1 Spontaneous
Egr3-/-
Drag
n.s.
o D nta ai n ly eo ag us on is ts
Egr3-/-
0.5
-0.5
40°
Hindlimb Ipsi stance Contra stance
Sp
Wild-type
Iliac crest Stance Swing
Hindlimb Ipsi stance Contra stance
C
Sp o D nta ai n ly eo ag u on s is ts
Daily agonists exposure
PC1 (25%): Recovery (a.u.)
Spontaneous
PC1 (25%): Recovery (a.u.)
B
r = 0.73
0.5 0 -0.5 -1 -1 -0.5 0 0.5 1 Daily agonist exposure
Figure 4. Impact of Activity on Functional Recovery after Spinal Cord Injury (A) Body trajectories measured during the first 5 min of home cage monitoring for wild-type (n = 4) and Egr3 mutant (n = 3) mice. Quantification of distance covered during 20 min period at intact condition and throughout recovery (p = 0.86; no significant effect of genotype). (B) Stick representation of hindlimb movement at the chronic stage during treadmill locomotion shown for spontaneous and daily agonist exposure groups (below: concurrent limb endpoint trajectories and velocity vector at swing onset, together with ipsilesional hindlimb oscillations; dark gray bars, stance; red bars, dragging). (C) Histogram plots reporting mean values of scores on recovery-related PC1 (25% explained variance) performed on ipsilesional gait patterns (average of 10–20 gait cycles/mouse, 108 parameters per gait cycle; n = 4 [agonist exposure] or 9 [spontaneous] wild-type and n = 5 [both conditions] Egr3 mutant mice). Within genotype, scores are not different between spontaneous and daily agonist exposure groups before injury and throughout recovery process. (D) Factor loadings (correlation of kinematic parameter and recovery-associated PC) of PC1 for all parameters (individual dots) of the two conditions (spontaneous recovery, daily agonist exposure) were correlated against each other for wild-type and Egr3 mutant mice. Strong positive correlation represents similar recovery process in both spontaneous and agonist exposure groups for both genotypes. Error bars, SEM; ns, not significant; acute, 3 days postinjury; chronic, 7 weeks postinjury. See also Figures S1 and S2.
gigantocellular, spinal vestibular, and red nucleus, as well as in M1 motor cortex (Figures 5B, 5C, and S5B). These findings reveal an absence of significant baseline differences between wild-type and Egr3 mutant mice, allowing direct comparison of descending projection neuron populations across genotypes after injury. At the acute stage, the majority of ipsilesional brainstem nuclei were not labeled. This depletion results from the disrupted access of ipsilaterally projecting brainstem nuclei to circuits below lesion. Lesion also disconnected contralaterally projecting descending pathways that decussate above lesion, e.g., leading to a lack of Rab-FPON neurons in M1 motor cortex and the red nucleus (Figures 5B, 5C, and S5B). In contrast, we detected 1632 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
a fraction of retrogradely marked spinal vestibular neurons residing in the ipsilesional brainstem (Figures 5B, 5C, and S5B). Axons of such neurons cross the midline above lesion, descend the spinal cord contralaterally, and establish collaterals crossing the midline a second time below lesion. We classified neurons with such axonal trajectories as dual midline-crossing projection neurons. At chronic stages, we detected substantial reorganization of ipsilesional brainstem pathways in wild-type mice. Magnocellular, gigantocellular, and pontine nuclei contained significantly more ipsilesional Rab-FPON neurons than at acute stages (Figures 5B, 5C, and S5B). The presence of retrogradely labeled neurons in the ipsilesional brainstem thus implies that their axons
A
Intact
B Ipsi
Chronic
Acute
Brainstem
Cervical
Wild-type
Contra
Thoracic Hx Rabies
Ipsi
Contra
Ipsi
Contra
Ipsi
Contra
Ipsi
Contra Ipsi
Contra
Ipsi
Contra
Mc Gi Pn
R SpVe Ve
Egr3-/-
Lumbar
Wild-type Egr3-/-
% ipsi-lesion rabies-positive neurons
C
SpVe Spinovestibular **
Mc Magnocellular
100
80
75
60
50
40
**
Gi Gigantocellular 80
**
*
** ***
Pn Pontine nucleus
**
60
20
20
0
0
0
100
60 40
40
25
R Red nucleus ***
*
*
75 50
20 0
25 0
Figure 5. Reduced Injury Responses in Brainstem Pathways of Egr3 Mutant Mice (A) Diagram illustrating rabies virus injection strategy to retrogradely label brainstem neurons with descending projections to ipsilesional lumbar spinal cord (yellow). Bottom: display of different brainstem nuclei (Esposito et al., 2014; Paxinos and Franklin, 2012): Mc, magnocellular nucleus; Gi, Gigantocellular nucleus; Pn, Pontine nucleus; R, red nucleus; SpVe, spinal vestibular nucleus; Ve, vestibular nucleus. (B) Top-down snapshots of 3D brainstem reconstructions in wild-type (top) and Egr3 mutant (bottom) mice at intact, acute, and chronic stages (ipsi- and contralesional halves of the reconstruction displayed separately; each neuron represented by single dot; for color-code, see A). (C) Quantification of brainstem reconstruction data (n = 3 each for intact and acute wild-type and Egr3 mutant, n = 4 each for chronic wild-type and Egr3 mutant) displaying percentage of ipsilesional neurons of entire rabies-marked respective subpopulation. *p < 0.05; **p < 0.01; ***p < 0.001; error bars, SEM; acute, 3 days postinjury; chronic, 7 weeks postinjury; Hx, hemisection. See also Figure S5.
cross the midline twice to establish novel dual midline-crossing pathways. In contrast, Egr3 mutant mice showed reduced levels of brainstem projection reorganization, with no significant differences in ipsilesional Rab-FPON neurons in magnocellular, gigantocellular, and pontine nuclei at chronic compared to acute stages (Figures 5B, 5C, and S5B). In contrast, we did not detect significant differences for neurons in the red nucleus between wild-type and Egr3 mutant mice. Together, these results suggest that functional muscle spindle feedback facilitates rearrangement of specific descending pathways from the brainstem, but interestingly, not all populations were affected equally. Next, we evaluated the effect of hemisection lesion on spinal projection neurons. We used rabies viruses to label neurons
through their axonal projections (Figures S6A–S6C). We also exploited a transsynaptic virus-based approach with monosynaptic restriction to capture synaptic connectivity (Wickersham et al., 2007) (Figure 6). Both approaches revealed similar distribution patterns of spinal projection neurons across multiple segments of the spinal cord at intact stages in both wild-type and Egr3 mutant mice (Figures 6 and S6A–S6C). At acute stages, only few ipsilesional spinal projection neurons exhibited dual midlinecrossing circuitry (Figures 6B and 6C; Figures S6B and S6C). In contrast, contralesional spinal neurons were abundantly marked by Rab-FP. At chronic stages, Rab-FP injections in wildtype mice revealed a prominent increase in the percentage of ipsilesional neurons compared to acute stages (Figures 6B and Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. 1633
A
TVA/G Rabies
Ipsi Contra
Starter neurons
B % ipsi-lesional rabies positive neurons *
60 40 20
AAV-flex -TVA/G+Cre
Hx
0
2 weeks
2
ct
ta
In
EnvA-RFP 1 2 rabies Viruses Wild-type Acute
C
Chronic
Intact
Wild-type
Egr3-/-
Egr3-/Acute
Chronic
Thoracic
Cervical
Intact
Ac ut e C hr on ic
Motor neuron ron p pools ools o ols s
1
Ipsi-lesional PNs Contra-lesional PNs
µ
-500 0 500 µm
-400
0
400
6C; Figures S6B and S6C). In Egr3 mutant mice, however, their percentage above lesion was significantly lower than in wildtype mice (Figures 6B and 6C; Figures S6B and S6C). In contrast, we did not detect distribution differences of spinal projection neurons between genotypes below lesion (Figures S6D and S6E). Together, these results demonstrate that Egr3 mutant mice exhibit a deficiency in the establishment of dual-crossing detour circuits involving multiple populations of descending projection neurons, whereas these are abundantly detected in wild-type mice. Spinal Projection Neurons Connect to Deprived Circuits by Distinct Mechanisms To gain insight into the cellular mechanisms responsible for the emergence of ipsilesional dual-crossing detour circuits after hemisection, we devised anterograde virus-mediated tracing experiments from supralesional spinal segments. We used injections of viruses that allow visualizing axons and synapses of specific subpopulations of descending projection neurons. This strategy also enabled evaluation of circuit reorganization from contralesional neurons that are less amenable to assess with retrograde tracing approaches due to partial persistence of projections to ipsilesional circuits after lesion. 1634 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
Figure 6. Detour Circuit Formation after Spinal Cord Injury Is Reduced in Egr3 Mutant Mice (A) Diagram illustrating monosynaptic rabies virus injection strategy to retrogradely mark descending spinal projection neurons with synaptic connections to ipsilesional neurons below lesion (injection at L2–L5; yellow; neurons with dotted axons, severed by injury; magenta, dual-crossing ipsilesional neurons). Top-left corner: example of triple-labeled (TVA/G/Rabies) neurons. Right: lowresolution view and reconstruction of triple-positive starter neurons of representative spinal cord section. (B) Quantification of percentage of ipsilesional rabies positive spinal projection neurons above lesion with connections to ipsilesional starter neurons (n = 3 each for intact and acute wild-type and Egr3 mutant; n = 4 for chronic wild-type; n = 5 for Egr3 mutant). (C) 3D reconstructions of supralesional spinal projection neurons with connection to ipsilesional lumbar circuits below lesion (yellow) in wild-type (left) and Egr3 mutant (right) mice at intact, acute, and chronic stages in top-down longitudinal view (top) and transverse section (below) view (filled triangle, lesion position; gray line, midline; magenta, ipsilesional neurons). *p < 0.05; error bars, SEM; acute, 3 days postinjury; chronic, 7 weeks postinjury; Hx, hemisection. See also Figure S6.
Ipsi-lesional PNs Contra-lesional PNs
We conducted a series of anterograde tracing experiments from ipsilesional or contralesional circuits above injury (Figure 7A) by unilateral coinjections of AAV-Tomato and AAV-SynGFP or AAVSynMyc to anterogradely track axonal trajectories and synaptic arborizations (Figures 7 and S7). In intact mice of both genotypes, cervical (C5–7) and thoracic (T7–8) neurons projected to lumbar levels bilaterally (Figure S7A). After injury, ipsilesional injections target commissural neurons with axonal tracts descending contralateral to cell body position, whereas contralesional injections target ipsilateral projection neurons with axonal tracts ipsilateral to cell body position (Figure 7A). We found that for both ipsi- and contralesional injections, descending axon tracts were present bilaterally in the white matter above lesion (Figure S7B). After hemisection, tracts only persisted on the contralesional side below lesion, demonstrating that axon collaterals at lumbar levels were exclusively derived from neurons projecting through contralesional tracts (Figures 7A and S7B). Next, we determined the frequency of midline-crossing axon collaterals below lesion at 2 weeks postlesion, the earliest possible tracing time point, and at chronic stages (Figures 7B and 7C). At 2 weeks postinjury, no difference was observed for ipsi- or contralesional populations and their midline-crossing frequency between wild-type and Egr3 mutant mice (Figure 7C). At chronic stages, Egr3 mutants exhibited a significantly reduced number of midline-crossing axons derived from ipsilesional populations compared to wild-type
Ipsi or contra AAV tdTomato
I
I
C
B
Transverse view
I
Above lesion
Midline-crossing axons Ipsi-lesional PNs
C
Ipsi-lesional PNs AAV Syn-Myc AAV nls-Tag1 Hx
Contra-lesional PNs
Below lesion ?
Contra-lesional PNs AAV Syn-GFP AAV nls-Tag2
C
iipsi psi 4 3 2 1 0 2
5 4 3 2 1 0
wk
s r ch
on
ic
Wild-type Egr3-/-
co ontra contra Egr3-/-
Wild-type
500
µm
D
*
Ipsi-lesional PNs
*
0
-500 -400
0 µm
400
500
µm
5
Contra-lesional PNs
3 Midline-crossing axons per 10 axon tracts
Ipsi-lesional PNs Contra-lesional PNs
Contra
?
?
C
Ipsi
Synaptic reconstruction
0
-500
Wild-type Egr3-/-
30 Ipsi-lesional synaptic terminals (%)
Longitudinal view
Co-localization Spot detection Synaptic tag
A
*
*
20 10 0 2
30
wk
s ch *
ro
ni
c
*
20 10 0
-400
0
400
Figure 7. Distinct Reconnection Mechanisms for Spinal Projection Neuron Subpopulations (A) Diagrams illustrating intraspinal injection scheme to anterogradely visualize axons (tdTomato) and synapses (synaptically tagged proteins) of ipsilesional (magenta) or contralesional (blue) spinal projection neuron residing above lesion (nuclear markers confirm unilaterality of injection). Top-down longitudinal and cross-section projected views shown (yellow, ipsilesional territory below lesion). (B) Examples of midline-crossing axons and synaptic terminal analysis with high-resolution imaging and Imaris spot detection. (C) Frequency analysis of midline-crossing axons originating from ipsi- and contralesional cervical spinal projection neurons, normalized to number of marked axons in contralesional white matter tracts below lesion (ipsilesion: n = 3 each for wild-type and Egr3 mutant for both time points; contralesion: n = 4 for wild-type and n = 3 Egr3 mutant for 2 week time point; n = 4 for wild-type and n = 5 Egr3 mutant for chronic analysis). (D) Quantitative analysis of distribution and density of synaptic terminals in the spinal cord below injury, originating from ipsi- and contralesional cervical spinal projection neurons (yellow, ipsilesional territory below lesion; ipsilesion, n = 4 for wild-type and n = 5 for Egr3 mutant for 2 week time point; n = 5 for wild-type and n = 6 for Egr3 mutant for chronic analysis; contra-lesion, n = 4 for wild-type and n = 6 for Egr3 mutant for 2 week time point; n = 5 each for wild-type and Egr3 mutant for chronic analysis). Contour plots show overall distribution of terminals from one chronic animal for each genotype; histogram plots display percentage of ipsilesional synaptic terminals at analyzed segmental spinal levels. *p < 0.05; error bars, SEM; Hx, hemisection; PN, projection neuron; chronic, 7 weeks postinjury. See also Figure S7.
mice, whereas no significant difference was observed for contralesional populations (Figures 7C and S7F). Together, these findings indicate that the absence of muscle spindle feedback impairs the ability to establish de novo dual midline-crossing axons originating from ipsilesional spinal projection neurons and that these anatomical differences between the two genotypes become apparent later than 2 weeks after injury. Next, we quantified ipsilesional synaptic arborization of midline-crossing axons (Figures 7B and 7D; Figures S7D and S7G). We reconstructed synaptic puncta at high resolution, yielding quantitative information on the spatial distribution and
number of synaptic terminals (Figure 7B). Analysis of ipsi- and contralesional projection neurons revealed comparable synaptic innervation above lesion between wild-type and Egr3 mutant mice (Figure S7D). In wild-type mice at chronic stages below lesion, synaptic input to ipsilesional gray matter targeted the ventrolateral quadrant, which contains many locomotor interneurons and motor neurons (Figures 7D and S7G). In contrast, the distribution of synaptic input beyond the midline in Egr3 mutant mice was primarily confined to medially located territory (Figures 7D and S7G). The observed increase in synaptic terminals in wild-type mice was not present 2 weeks postlesion, Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. 1635
in agreement with the corresponding time course of midlinecrossing axon elaboration (Figures 7C and 7D). For contralesional projection neurons, we also detected lower synaptic terminal density in the ipsilesional gray matter below lesion in Egr3 mutant mice at chronic stages compared to wildtype mice, despite a similar number of midline-crossing axons in both genotypes (Figures 7C and 7D; Figures S7D and S7G). Strikingly, however, we found a selective decrease in the density of synaptic terminals between 2 weeks postlesion and chronic stages in Egr3 mutant mice, ultimately leading to the observed lower terminal density compared to wild-type mice (Figure 7D). Together, our findings provide evidence that after lateral hemisection spinal cord injury, muscle spindle feedback enhances the process of axonal and synaptic rearrangements of multiple descending spinal projection neuron populations through distinct mechanisms. DISCUSSION Spinal cord injuries lead to immediate motor dysfunction because of separation of descending control pathways from local spinal circuits. Various degrees of functional recovery occur after incomplete injury. However, the likely involvement of numerous circuit elements paired with the limited understanding of their precise organization and function within the hierarchy of motor control pathways have posed challenges for gaining mechanistic insight in the process of functional recovery. Here, we demonstrate that muscle spindle feedback circuits are essential to direct locomotor recovery after lateral hemisection spinal cord injury and that the lack of this specific sensory channel affects the ability of descending projection neurons to undergo efficient circuit reorganization after injury. We discuss our findings with an emphasis on the role of sensory feedback circuits in locomotor improvement after injury and the mechanisms by which circuit rearrangements parallel and influence the recovery process. Task-Specific Locomotor Recovery after Spinal Cord Injury in Wild-Type Mice Wild-type mice improve basic locomotor function after hemisection spinal cord injury to a significant extent. In contrast, they remain severely compromised in their ability to carry out precision ladder walking. These findings underscore the need for task-specific communication channels between supraspinal and spinal circuits, some of which do not recover after injury. A possible model to explain these findings is that upon establishment of multistep synaptic relays, a comparatively crude wiring of descending circuit elements is sufficient to drive disconnected ipsilesional spinal circuits below lesion for regaining basic locomotor function. Newly established descending connections can interact with an already wired repertoire of local spinal circuits able to coordinate basic locomotor behaviors. In contrast, precision tasks likely require specific and refined descending circuit connectivity. In addition, complex tasks may depend more heavily on information conveyed through ascending pathways, which exhibit enduring dysfunction after spinal cord injury (Kaas et al., 2008; Martinez et al., 2010). Taken together, these observations suggest that distinct neuronal circuit ele1636 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
ments are responsible and necessary for the re-establishment of task-specific functions. Role of Muscle Spindle Feedback Circuits in Locomotor Recovery after Spinal Cord Injury Muscle spindle afferents constitute a minor fraction of DRG sensory neurons (Scott, 1992), but our results demonstrate that they are essential to promote locomotor recovery after incomplete spinal cord lesion. Why does deprivation of a specific sensory channel lead to such profound impairment? Each class of functionally distinct sensory neurons exhibits lamina-specific axonal terminations in the spinal cord (Brown, 1981). While cutaneous and mechanoreceptive afferents target dorsal horn neurons, proprioceptive afferents terminate more ventrally, raising the possibility that these differential synaptic connectivity profiles may contribute to their role in the recovery process. A primary mode of action by muscle spindle afferents in facilitating recovery may involve recruitment of motor circuits through their unique connections. Targeted circuit elements include motor neurons and core components of ventral locomotor interneuron circuits that have recently been demonstrated to play important roles in the regulation of extensor-flexor alternation (Talpalar et al., 2011; Zhang et al., 2014) and rhythm generation (Dougherty et al., 2013) in the mouse. The pivotal role of muscle spindle feedback in promoting locomotor improvement after lateral hemisection observed here might therefore be at least in part attributed to their direct synaptic access to these neurons. Specifically, muscle spindle afferents are embedded in a highly selective central synaptic connectivity matrix. Transfer of muscle-specific information to functionally distinct interneurons that directly activate motor neurons or mediate reciprocal inhibition between motor neurons is a key feature of these neuronal networks (Jankowska and Edgley, 1993; McCrea and Rybak, 2008; Pearson, 2008; Wang et al., 2008; Windhorst, 2007). Muscle spindle afferent recruitment after injury may strengthen these specific spinal circuits and their connections (Petruska et al., 2007), whereas their functional absence in Egr3 mutants might contribute to the severe impairment in recovery. An alternative or complementary possibility is that muscle spindle afferents release factors in an activity-dependent manner, which in turn promote circuit reorganization in the spinal cord. For instance, retrograde trophic support by the neurotrophin NT-3 strengthens synaptic connections (Boyce and Mendell, 2014; Chen et al., 2002; Oakley et al., 1997). Moreover, the amount of physical activity influences baseline BDNF expression in the spinal cord after traumatic injury (Ying et al., 2008), an effect that may be mediated by recruitment of muscle spindle feedback circuits. In our experiments, we found no difference in the degree of spontaneous cage activity between wildtype and Egr3 mutants after hemisection spinal cord injury, excluding disparity in physical activity as a possible reason for differential recovery. In addition, daily application of monoaminergic agents to enhance activity of local spinal circuits in an attempt to bypass reduced sensory feedback in Egr3 mutant mice was inefficient in overcoming the severely limited recovery in Egr3 mutant mice. These findings demonstrate that muscle spindle afferents, despite being a numerically minor sensory
neuron population, play an instrumental and selective role in promoting functional recovery after spinal cord injury. Formation of Spinal Detour Circuits Parallels Locomotor Recovery Regaining locomotor function of the ipsilesional hindlimb after thoracic hemisection requires the establishment of detour circuits that reconnect descending pathways to deprived locomotor circuits below lesion. The formation of such detour circuits to functionally bridge the injury site depends on local axon growth and reorganization of synaptic connectivity within existing descending circuit modules (Ballermann and Fouad, 2006; Bareyre et al., 2004; Courtine et al., 2008; Jankowska and Edgley, 2006; Rosenzweig et al., 2010; van den Brand et al., 2012). We demonstrate that in wild-type mice, injury-induced circuit-level responses involve the deployment of specific patterns of axonal growth and synaptic arborization from distinct populations of supraspinal and spinal projection neurons. Our anatomical mapping to identify injury-responsive descending circuit elements above lesion demonstrates that reduced compensatory responses to injury are widespread in Egr3 mutants. These alterations include ipsi- and contralesional spinal projection neurons at multiple spinal segments and specific descending pathways from the brainstem. Perturbation or silencing of any identified specific neuron population alone in wild-type mice is therefore unlikely to recapitulate the dramatic lack of recovery observed in Egr3 mutant mice. On the other hand, experimental attempts to specifically target a majority of neurons undergoing novel collateral formation after injury would require injections at multiple central nervous system (CNS) sites, likely themselves inducing behavioral repercussions. In addition, even if successful, such approaches would interfere with the function of targeted neurons in their entirety, and not just with the newly formed collaterals. How does muscle spindle feedback facilitate de novo circuit formation? While we cannot rule out multifaceted circuit-level effects influencing the recovery process, we favor a model in which muscle spindle feedback circuits act primarily on ipsilesional circuits below the injury site to promote the formation of compensatory connections to deprived circuits. In agreement with such a model, identified brainstem populations do not receive direct synaptic input from muscle spindle afferents, implying that at least for these populations such input is not essential to trigger circuit reorganization. Mechanistically, the assembly of novel circuits in the adult nervous system may be achieved through stabilization of nascent axon collaterals involving Hebbian plasticity reinforced by muscle spindle afferent input. Growth and stabilization of axons in the developing nervous system suggests that such mechanisms act in highly cell-type-specific patterns (Andreae and Burrone, 2014). To gain insight into how defined neuronal populations respond to injury, we focused our anterograde synaptic mapping analysis on spinal projections neurons. Comparison of wild-type and Egr3 mutant mice uncovered distinct responses for specific spinal populations. Ipsilesional descending spinal projection neurons in Egr3 mutants exhibited both a reduction in dual midline-crossing axons and decreased ipsilesional synaptic arborization below lesion, occurring later than 2 weeks
after injury. In contrast, contralesional counterparts only showed restricted arborization of synaptic terminals without disruption in midline-crossing axons. These synaptic differences however can be attributed to synaptic pruning in the absence of muscle spindle feedback rather than additional synaptic growth in wild-type mice. These findings also suggest that the majority of injury-responsive contralesional spinal projection neurons already possess midline-crossing collaterals at intact stages, providing an explanation for why this parameter is not affected in Egr3 mutants compared to wild-type mice at chronic stages. In summary, our study demonstrates that one specific sensory channel has an executive role in directing restoration of hindlimb motor function and facilitating multifaceted circuit reorganization after incomplete spinal cord injury. These findings stress the importance of exploiting muscle spindle feedback circuits in the design of rehabilitative strategies after spinal cord injury. Epidural stimulation of lumbar segments facilitates motor control and leads to improved functional recovery in animal models and paraplegic individuals (Angeli et al., 2014; van den Brand et al., 2012). This treatment paradigm may at least in part act through the recruitment of myelinated sensory feedback circuits (Capogrosso et al., 2013). Refined experimental strategies to specifically modulate muscle spindle feedback channels open innovative therapeutic avenues to pursue in the future. Similar concepts may apply to other traumatic CNS disorders, such as stroke or brain injury, which heavily rely on plasticity of both supraspinal and spinal descending pathways to regain functional capacities after lesion. EXPERIMENTAL PROCEDURES Mouse Genetics and Surgeries Mice used were from a local colony containing the Egr3 mutant allele previously described (Tourtellotte and Milbrandt, 1998). Surgical procedures for hemisection injury and EMG implantation have been described previously (Courtine et al., 2008) and were performed under full general anesthesia with isoflurane in oxygen-enriched air (1%–2%). Local Swiss veterinary offices approved all the procedures. Details on mice and surgical procedures are described in the Extended Experimental Procedures. Behavioral Analysis Whole-body kinematics were recorded using the high-speed motion capture system Vicon (Vicon Motion Systems), combining 10–12 infrared cameras (200 Hz) (van den Brand et al., 2012). Parameters describing kinematic and EMG characteristics were computed using custom-written MATLAB scripts (van den Brand et al., 2012). Behavioral tests included overground locomotion on an elevated runway, stepping on a motorized treadmill (Robomedica), elevated horizontal ladder, and swimming. To quantify task- and genotypespecific gait characteristics prior to injury and throughout the recovery process after hemisection spinal cord injury, we implemented a multistep statistical procedure based on PC analysis (Dominici et al., 2012). A flowchart explaining the various steps of this analysis can be found in Figure S2. For behavioral monitoring of home cage activity, spontaneous activity was surveyed for each mouse during 20 min. Additional information on recordings, postprocessing and behavioral tasks are available in the Extended Experimental Procedures and Figure S2. Anatomical Tracing Experiments Rabies viruses and AAVs were amplified and purified from local viral stocks following established protocols (Esposito et al., 2014; Pivetta et al., 2014; Wickersham et al., 2010). Additional information on production and injection
Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. 1637
of viruses, antibodies, imaging, and anatomical quantification can be found in the Extended Experimental Procedures. Statistical Analysis All data are reported as mean values ± SEM. All statistical evaluations were performed using GraphPad Prism (v. 6.0) (Prism, GraphPad Software) using unpaired Student’s t test (Figures 1C, 1D, 2D, 5C, 6B, 7C, and 7D; Figures S3A, S3B, S4D, S6C, S7A, S7B, S7D, S7F, and S7G), two-way ANOVA for repeated measurement (Figures 1D, 3D, 3E, 4A, and 4C; Figures S3D, S3E, S5A, and S5B), and one-way ANOVA for repeated measurements (Figure S5C), ´da´k-Bonferroni). The significance level followed by post hoc comparisons (Sı for behavioral analysis was set as jR valuej > 0.5 and p < 0.05, respectively. Significance level is defined as follows for all analyses performed: *p < 0.05; **p < 0.01; ***p < 0.001. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven figures, and five movies and can be found with this article online at http://dx. doi.org/10.1016/j.cell.2014.11.019. AUTHOR CONTRIBUTIONS A.T., G.C., and S.A. initiated the project. A.T. and I.V. performed behavioral experiments and analysis; A.T. performed anatomy experiments and analysis. All authors were involved in interpretation of experiments and contributed to writing the paper. ACKNOWLEDGMENTS We are grateful to M. Sigrist, M. Mielich, P. Capelli, and S. Esposito for expert technical help, M. Kirschmann, S. Bourke, and L. Gelman from the FMI imaging facility, N. Ehrenfurchter from the Biozentrum Imaging facility for help and advice with image acquisition and analysis, and P. Caroni for discussions and comments on the manuscript. A.T. and S.A. were supported by an ERC Advanced Grant, the Swiss National Science Foundation, the Kanton Basel-Stadt, and the Novartis Research Foundation. I.V. was supported by the European Neuroscience Campus (ENC) network. G.C. was supported by an ERC Starting Grant and the International Foundation for Research in Paraplegia (IRP). Received: August 13, 2014 Revised: November 5, 2014 Accepted: November 11, 2014 Published: December 18, 2014 REFERENCES Andreae, L.C., and Burrone, J. (2014). The role of neuronal activity and transmitter release on synapse formation. Curr. Opin. Neurobiol. 27, 47–52. Angeli, C.A., Edgerton, V.R., Gerasimenko, Y.P., and Harkema, S.J. (2014). Altering spinal cord excitability enables voluntary movements after chronic complete paralysis in humans. Brain 137, 1394–1409. Arshavskiĭ, IuI., Kots, IaM., Orlovskiĭ, G.N., Rodionov, I.M., and Shik, M.L. (1965). [Study of biomechanics of running dogs]. Biofizika 10, 665–671. Ballermann, M., and Fouad, K. (2006). Spontaneous locomotor recovery in spinal cord injured rats is accompanied by anatomical plasticity of reticulospinal fibers. Eur. J. Neurosci. 23, 1988–1996. Bareyre, F.M., Kerschensteiner, M., Raineteau, O., Mettenleiter, T.C., Weinmann, O., and Schwab, M.E. (2004). The injured spinal cord spontaneously forms a new intraspinal circuit in adult rats. Nat. Neurosci. 7, 269–277.
Brown, A.G. (1981). Organization of the Spinal Cord (New York: Springer Verlag). Capogrosso, M., Wenger, N., Raspopovic, S., Musienko, P., Beauparlant, J., Bassi Luciani, L., Courtine, G., and Micera, S. (2013). A computational model for epidural electrical stimulation of spinal sensorimotor circuits. J. Neurosci. 33, 19326–19340. Chen, H.H., Tourtellotte, W.G., and Frank, E. (2002). Muscle spindle-derived neurotrophin 3 regulates synaptic connectivity between muscle sensory and motor neurons. J. Neurosci. 22, 3512–3519. Courtine, G., Song, B., Roy, R.R., Zhong, H., Herrmann, J.E., Ao, Y., Qi, J., Edgerton, V.R., and Sofroniew, M.V. (2008). Recovery of supraspinal control of stepping via indirect propriospinal relay connections after spinal cord injury. Nat. Med. 14, 69–74. Curt, A., Van Hedel, H.J., Klaus, D., and Dietz, V.; EM-SCI Study Group (2008). Recovery from a spinal cord injury: significance of compensation, neural plasticity, and repair. J. Neurotrauma 25, 677–685. Dietz, V., and Fouad, K. (2014). Restoration of sensorimotor functions after spinal cord injury. Brain 137, 654–667. Dougherty, K.J., Zagoraiou, L., Satoh, D., Rozani, I., Doobar, S., Arber, S., Jessell, T.M., and Kiehn, O. (2013). Locomotor rhythm generation linked to the output of spinal shox2 excitatory interneurons. Neuron 80, 920–933. Eccles, J.C., Eccles, R.M., and Lundberg, A. (1957). The convergence of monosynaptic excitatory afferents on to many different species of alpha motoneurones. J. Physiol. 137, 22–50. Edgerton, V.R., Courtine, G., Gerasimenko, Y.P., Lavrov, I., Ichiyama, R.M., Fong, A.J., Cai, L.L., Otoshi, C.K., Tillakaratne, N.J., Burdick, J.W., and Roy, R.R. (2008). Training locomotor networks. Brain Res. Brain Res. Rev. 57, 241–254. Esposito, M.S., Capelli, P., and Arber, S. (2014). Brainstem nucleus MdV mediates skilled forelimb motor tasks. Nature 508, 351–356. Gruner, J.A., and Altman, J. (1980). Swimming in the rat: analysis of locomotor performance in comparison to stepping. Exp. Brain Res. 40, 374–382. Halbertsma, J.M. (1983). The stride cycle of the cat: the modelling of locomotion by computerized analysis of automatic recordings. Acta Physiol. Scand. Suppl. 521, 1–75. Jankowska, E., and Edgley, S. (1993). Interactions between pathways controlling posture and gait at the level of spinal interneurones in the cat. Prog. Brain Res. 97, 161–171. Jankowska, E., and Edgley, S.A. (2006). How can corticospinal tract neurons contribute to ipsilateral movements? A question with implications for recovery of motor functions. Neuroscientist 12, 67–79. Kaas, J.H., Qi, H.X., Burish, M.J., Gharbawie, O.A., Onifer, S.M., and Massey, J.M. (2008). Cortical and subcortical plasticity in the brains of humans, primates, and rats after damage to sensory afferents in the dorsal columns of the spinal cord. Exp. Neurol. 209, 407–416. Knikou, M., and Mummidisetty, C.K. (2014). Locomotor training improves premotoneuronal control after chronic spinal cord injury. J. Neurophysiol. 111, 2264–2275. Lavrov, I., Courtine, G., Dy, C.J., van den Brand, R., Fong, A.J., Gerasimenko, Y., Zhong, H., Roy, R.R., and Edgerton, V.R. (2008). Facilitation of stepping with epidural stimulation in spinal rats: role of sensory input. J. Neurosci. 28, 7774–7780. Maier, I.C., and Schwab, M.E. (2006). Sprouting, regeneration and circuit formation in the injured spinal cord: factors and activity. Philos. Trans. R. Soc. Lond. B Biol. Sci. 361, 1611–1634. Martinez, M., Delcour, M., Russier, M., Zennou-Azogui, Y., Xerri, C., Coq, J.O., and Brezun, J.M. (2010). Differential tactile and motor recovery and cortical map alteration after C4-C5 spinal hemisection. Exp. Neurol. 221, 186–197.
Bouyer, L.J., and Rossignol, S. (2003). Contribution of cutaneous inputs from the hindpaw to the control of locomotion. II. Spinal cats. J. Neurophysiol. 90, 3640–3653.
McCrea, D.A., and Rybak, I.A. (2008). Organization of mammalian locomotor rhythm and pattern generation. Brain Res. Brain Res. Rev. 57, 134–146.
Boyce, V.S., and Mendell, L.M. (2014). Neurotrophins and spinal circuit function. Front. Neural Circuits 8, 59.
Oakley, R.A., Lefcort, F.B., Clary, D.O., Reichardt, L.F., Prevette, D., Oppenheim, R.W., and Frank, E. (1997). Neurotrophin-3 promotes the differentiation
1638 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
of muscle spindle afferents in the absence of peripheral targets. J. Neurosci. 17, 4262–4274.
Tourtellotte, W.G., and Milbrandt, J. (1998). Sensory ataxia and muscle spindle agenesis in mice lacking the transcription factor Egr3. Nat. Genet. 20, 87–91.
Paxinos, G., and Franklin, K.B. (2012). The Mouse Brain in Stereotaxic Coordinates, Fourth Edition (San Diego: Elsevier).
van den Brand, R., Heutschi, J., Barraud, Q., DiGiovanna, J., Bartholdi, K., Huerlimann, M., Friedli, L., Vollenweider, I., Moraud, E.M., Duis, S., et al. (2012). Restoring voluntary control of locomotion after paralyzing spinal cord injury. Science 336, 1182–1185.
Pearson, K.G. (2008). Role of sensory feedback in the control of stance duration in walking cats. Brain Res. Brain Res. Rev. 57, 222–227. Petruska, J.C., Ichiyama, R.M., Jindrich, D.L., Crown, E.D., Tansey, K.E., Roy, R.R., Edgerton, V.R., and Mendell, L.M. (2007). Changes in motoneuron properties and synaptic inputs related to step training after spinal cord transection in rats. J. Neurosci. 27, 4460–4471. Pivetta, C., Esposito, M.S., Sigrist, M., and Arber, S. (2014). Motor-circuit communication matrix from spinal cord to brainstem neurons revealed by developmental origin. Cell 156, 537–548. Ramon y Cajal, S. (1928). Degeneration and Regeneration of the Nervous System (Oxford: Oxford University Press). Rosenzweig, E.S., Courtine, G., Jindrich, D.L., Brock, J.H., Ferguson, A.R., Strand, S.C., Nout, Y.S., Roy, R.R., Miller, D.M., Beattie, M.S., et al. (2010). Extensive spontaneous plasticity of corticospinal projections after primate spinal cord injury. Nat. Neurosci. 13, 1505–1510. Rossignol, S., and Frigon, A. (2011). Recovery of locomotion after spinal cord injury: some facts and mechanisms. Annu. Rev. Neurosci. 34, 413–440. Rossignol, S., Dubuc, R., and Gossard, J.P. (2006). Dynamic sensorimotor interactions in locomotion. Physiol. Rev. 86, 89–154. Roy, R.R., Harkema, S.J., and Edgerton, V.R. (2012). Basic concepts of activity-based interventions for improved recovery of motor function after spinal cord injury. Arch. Phys. Med. Rehabil. 93, 1487–1497. Scott, S.A. (1992). Sensory Neurons: Diversity, Development and Plasticity (New York: Oxford University Press). Talpalar, A.E., Endo, T., Lo¨w, P., Borgius, L., Ha¨gglund, M., Dougherty, K.J., Ryge, J., Hnasko, T.S., and Kiehn, O. (2011). Identification of minimal neuronal networks involved in flexor-extensor alternation in the mammalian spinal cord. Neuron 71, 1071–1084. Tello, F. (1907). La regeneration dans les voies optiques. Trab. Lab. Invest. Biol. Univ. Madr. 5, 237–248.
Wang, Z., Li, L., Goulding, M., and Frank, E. (2008). Early postnatal development of reciprocal Ia inhibition in the murine spinal cord. J. Neurophysiol. 100, 185–196. Wickersham, I.R., Sullivan, H.A., and Seung, H.S. (2010). Production of glycoprotein-deleted rabies viruses for monosynaptic tracing and high-level gene expression in neurons. Nat. Protoc. 5, 595–606. Wickersham, I.R., Lyon, D.C., Barnard, R.J., Mori, T., Finke, S., Conzelmann, K.K., Young, J.A., and Callaway, E.M. (2007). Monosynaptic restriction of transsynaptic tracing from single, genetically targeted neurons. Neuron 53, 639–647. Windhorst, U. (2007). Muscle proprioceptive feedback and spinal networks. Brain Res. Bull. 73, 155–202. Ying, Z., Roy, R.R., Zhong, H., Zdunowski, S., Edgerton, V.R., and Gomez-Pinilla, F. (2008). BDNF-exercise interactions in the recovery of symmetrical stepping after a cervical hemisection in rats. Neuroscience 155, 1070–1078. Zhang, J., Lanuza, G.M., Britz, O., Wang, Z., Siembab, V.C., Zhang, Y., Velasquez, T., Alvarez, F.J., Frank, E., and Goulding, M. (2014). V1 and v2b interneurons secure the alternating flexor-extensor motor activity mice require for limbed locomotion. Neuron 82, 138–150. Zo¨rner, B., Filli, L., Starkey, M.L., Gonzenbach, R., Kasper, H., Ro¨thlisberger, M., Bolliger, M., and Schwab, M.E. (2010). Profiling locomotor recovery: comprehensive quantification of impairments after CNS damage in rodents. Nat. Methods 7, 701–708. Zo¨rner, B., Bachmann, L.C., Filli, L., Kapitza, S., Gullo, M., Bolliger, M., Starkey, M.L., Ro¨thlisberger, M., Gonzenbach, R.R., and Schwab, M.E. (2014). Chasing central nervous system plasticity: the brainstem’s contribution to locomotor recovery in rats with spinal cord injury. Brain 137, 1716–1732.
Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. 1639
Supplemental Information EXTENDED EXPERIMENTAL PROCEDURES Mouse Genetics and Maintenance Mice used in this study were derived from a colony containing the Egr3 mutant allele previously described (Tourtellotte and Milbrandt, 1998). They were maintained as heterozygous breeding pairs on a mixed genetic background (129/C57BL6). Wild-type and Egr3 mutant mice were identified by genotyping and subjected to experimental procedures described below, using both males and females without discernable differences in experimental outcome. Mice were housed on a 12 hr light/dark cycle with ad libitum access to food and water. They were handled for one week prior to the first surgeries. Animal care, including manual bladder voiding, was performed twice daily for 1 week after injury and as needed for the remaining postinjury period. Surgical Procedures To access the spinal cord for hemisection injury, it was exposed by a mid-dorsal skin incision and a laminectomy was made over spinal segment T10. A unilateral hemisection was performed using micro-scissors. After surgery, mice were placed in an incubator for optimized recovery from anesthesia. Prior to hemisection surgery, 20 mice (n = 12 wild-type and n = 8 Egr3 mutants) underwent implantation of bipolar electrodes into selected leg muscles to record EMG activity. Electrodes were implanted bilaterally into the medial gastrocnemius or vastus lateralis and into tibialis anterior. A common ground wire (AS631-2, Coonerwires) was inserted subcutaneously in the shoulder area and bipolar intramuscular EMG electrodes were inserted bilaterally in the mid-belly of the target muscles. Wires were routed subcutaneously through the back to a small percutaneous amphenol connector (Omnetics Connector Corporation, USA) securely cemented to the skull of the mouse. After a recovery time of 7 days, pre-lesion data was recorded as described below. Following pre-lesion data acquisition, mice underwent a second surgery during which the lateral hemisection was placed as described previously (Courtine et al., 2008). Analgesia (buprenorphine Temgesic, ESSEX Chemie AG, Switzerland, 0.05-0.1 mg per kg, s.c.) and antibiotics (Baytril 2,5%, Bayer Health Care AG, Germany, 5 10 mg per kg, s.c.) were provided for 3 and 5 days postsurgery, respectively. Kinematic Recordings All procedures used have been detailed previously (Dominici et al., 2012). Briefly, whole-body kinematics were recorded using the high-speed motion capture system Vicon (Vicon Motion Systems, UK), combining 10-12 infrared cameras (200 Hz). 3mm reflective markers were attached bilaterally overlying the iliac crest, the greater trochanter (hip), the lateral condyle (knee), the malleolus (ankle), and the base of the metatarsal phalageal joint (MTP) for the hindlimbs; and the proximal head of the humerus (shoulder), epicondyle of humerus (elbow) and on the medial head of metacarpal (forepaw) for the forelimbs. 3D position of the markers was reconstructed offline using Vicon Nexus software (1.4–1.8). The body was modeled as an interconnected chain of rigid segments (Movies S1, S2, S3, S4, and S5), and joint angles were generated accordingly. For subsequent kinematic analysis, only hindlimb and parameters related to the trunk were analyzed. Parameters (Figure S1) describing gait timing, joint kinematics, limb endpoint trajectory, and trunk stability were computed for each gait cycle using custom written MATLAB scripts and according to methods described previously (Dominici et al., 2012). Hindlimb kinematics during swimming were captured by two cameras (100 Hz; Basler Vision Technologies, Germany). 3mm markers were painted with a bright fluorescent color and placed over the iliac crest and on the base of the second metatarsus. For analysis, hindlimb movements were reconstructed as a single segment connecting the iliac crest and the MTP. Kinematracer (Kissei Comtec Co., Japan) motion tracking software was used to obtain 2D coordinates of joint positions for offline analysis of kinematic parameters related to the hindlimb (interlimb coordination) and the endpoint. EMG Recordings EMG signals were amplified using a differential AC amplifier 1700 (A-M Systems, Carlsborg, WA, USA), filtered, stored and analyzed off-line using custom written MATLAB script to compute co-contraction between antagonist muscles, and the amplitude, duration, and timing of individual bursts (Figure S1). Behavioral Tasks Prior to the first surgery, mice were trained for 5 days/week during 2 weeks in order to accustom them to the different locomotor tasks. After hemisection injury, mice were trained twice per week in addition to the weekly recordings in order to maintain learned proficiency of locomotor tasks tested. Overground locomotion was tested on a narrow (5cm wide, 1 m long), elevated runway. Mice were recorded at regular intervals throughout the recovery period after hemisection spinal cord injury (3 days, 1, 2, 4 and 7 weeks postinjury). 10-25 steps were analyzed per mouse and time point studied. For analysis, stepping velocity was matched as closely as possible in order to extract genotype-specific or injury-specific gait alterations. For this, only gait cycles with mean body velocities between 8-13 cm/s were considered for analysis. A total of 103 kinematic parameters were computed and included in the subsequent PC analysis for this task (Figure S1). Mice were also tested during stepping on a motorized treadmill (Robomedica, Inc., Mission Viejo, CA, USA). The speed of the treadmill belt was set at values ranging from 7-23 cm/s, with increments of 4 cm/s. After receiving a hemisection, mice were recorded at regular intervals throughout the recovery period (3 days, 1, 2, 4 and 7 weeks post injury). For both wild-type and Egr3 mutant mice, Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. S1
4 animals received chronic EMG implantation in the medial gastrocnemius and tibialis anterior muscle in order to study modulation of muscle activity with increasing treadmill speeds at intact and chronic states. Only wild-type mice were capable to step across the entire range of speeds at chronic stages after hemisection spinal cord injury. 10-30 steps were extracted per mouse analyzed. A total of 108 kinematic parameters were computed and included in the subsequent PC analysis for this task (Figure S1). Crossing of an elevated horizontal ladder (1 m long, rung diameter 5 mm, spaced at 2 cm intervals) was recorded pre-lesion and 7 weeks (chronic) after hemisection spinal cord injury. 30-40 steps were analyzed per mouse and time point (n = 9 for both genotypes). Quantification of hit, slip or miss paw placement was assessed from slow motion videos acquired at 100 Hz. All rungs of the ladder were labeled with reflective markers. 3D position of the rungs was computed using Vicon Nexus. Horizontal paw placement with respect to the rungs was computed as the relative horizontal distance of the foot at the vertical level of the ladder rungs with respect to two consecutive rungs (analyzed rung position at 0%; next rungs placed at ± 100% respectively; see Figure 2B for scheme). Intact mice were tested in a swimming pool made of Plexiglas. The pool (10 cm wide) was filled up to 15 cm with water (temperature: 28-30 C). Mice were encouraged to swim toward an escape platform located 1 m ahead of them. Swimming performance was recorded using intact (n = 8 wild-type and n = 7 Egr3 mutant mice) and chronic (n = 4 wild-type mice) mice. A total of 49 or 36 (without EMG for chronic mice) kinematic parameters were computed and included in the subsequent PC analysis for this task (Figure S1). Monitoring Activity in Home Cage Enrichment toys and food pellets were removed from the home cage. A camera (DCS-2230, 25 Hz, D-Link GmbH, Eschborn, Germany) was positioned above the cage to monitor whole-body movement of the mice. Spontaneous activity was monitored for each mouse during 20 min, in a random order. Testing was performed twice before the hemisection injury, and at regular intervals throughout the recovery process after spinal cord injury. All monitoring was performed during mornings. Data was stored and analyzed offline using custom written MATLAB scripts. Briefly, videos were converted into a high contrast black and white format and the X and Y positions of the center of the mice were computed for every individual frame. Distance and velocity were computed based on these coordinates. Principal Component Analysis Behavioral data was analyzed by PC analysis (van den Brand et al., 2012). A flowchart explaining the various steps of this analysis can be found in Figure S2. Step 1: Continuous locomotion is recorded using a high-resolution kinematic system. Step 2: Custom-written MATLAB scripts are applied to reconstructed kinematic data in order to compute basic parameters and highly elaborated variables that provide a comprehensive quantification of gait features. All variables computed for each task are specified in Figure S1. Approximately 10 steps were extracted per mouse and experimental condition. Step 3: All computed variables were averaged for each mouse independently. The matrix combining all mean values of variables from all mice of an analysis was then subjected to a PC analysis. For this purpose, we used the correlation method, which adjusts the mean of the data to 0 and the SD to 1. This method allows the comparison of variables with disparate values (large versus small), and/or different variances. The result of the PC analysis is a new set of synthetic uncorrelated variables, i.e., the PCs, which each explains the maximum possible amount of variance. Step 4: The new coordinates of gait patterns along each PC, termed PC scores, are extracted for each mouse. PC scores are used to represent gait patterns in the ‘‘denoised’’ PC space to visualize differences between mice and experimental conditions. A 95% confidence elliptic fitting is applied on data points corresponding to a given experimental condition to highlight disparities between genotypes, tasks, and/or time points postlesion. Step 5: PC scores are averaged for each experimental condition and represented in histogram plots to identify the type of information differentiated along each PC axis. Step 6: Each PC is a linear combination of the original parameters with appropriate weights, which are termed ‘‘factor loadings.’’ The values of factor loadings range from 1 to 1, and correspond to correlations between original parameters and a given PC. Step 7: Factor loadings with a high value (jfactor loadingj > 0.5) are extracted, color-coded based on their correlation value, and regrouped into functional clusters based on the type of gait control aspects they refer to. This process leads to an objective extraction of the most relevant behavioral parameters to account for the effects of a specific experimental condition. Step 8: To provide a more classical representation of the observed effects, relevant parameters representative of functional clusters are extracted and represented in histogram plots.
Monoaminergic Agonist Administration To enhance the activity of spinal interneurons after injury, mice were administered a daily injection of monoaminergic agonists, consisting of 5-HT2A/C agonist Quipazine (0.375 mg/kg), 5-HT1A/7 agonist (R)-(+)-8-Hydroxy-DPAT hydrobromide (0.125 mg/kg), DA1-like receptor agonist R(+)-SKF-81297 hydrobromide (0.3 mg/kg), delivered i.p. starting from day 3 postlesion ( = acute) to chronic stages. Behavioral recordings of the drug-administered cohort were conducted at the same time points after injury as the non-drug group (Figure 3A). Agonists at administered concentrations did not alter locomotor pattern of intact mice (data not shown). S2 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
Viral Tools and Production Rabies viruses used were amplified and purified from local viral stocks following established protocols (Rab-mCherry, Rab-GFP, and EnvA coated versions) (Esposito et al., 2014; Wickersham et al., 2010). AAVs used in this study were designed for conditional Cremediated expression. Therefore, they were all co-injected with AAV-Cre to allow for expression. All AAVs used in this study were previously described (Esposito et al., 2014; Pivetta et al., 2014), with the exception of a Synaptophysin-10xMyc-tag version constructed in analogy to the previously described Synaptophysin-GFP construct (Pivetta et al., 2014). AAV production (serotype 2.9) followed standard procedures and all AAVs used were of genomic titers > 1x10e13. Intraspinal Injections Using isoflurane anesthesia, mice underwent laminectomy to expose the surface of the spinal cord. For both retrograde and anterograde viral delivery into the spinal cord, a pulled calibrated glass pipette (Drummond Scientific) was used for local application of 100nl virus by multiple short pulses (3msec, 0.5Hz) using a picospritzer (Parker). After surgery, mice were administered Metacam (5 mg/kg) subcutaneously for analgesia. To verify injection precision and efficiency of infection, all mice were co-injected with AAVnuclear tags. Two weeks postvirus transduction, mice were sacrificed and unilaterality of injections was confirmed by immunohistochemistry. For lumbar-spinal cord initiated transsynaptic rabies experiments, we first performed unilateral co-injections of AAVs expressing G protein and TVA, and injected EnvA coated rabies two weeks subsequently as previously described (Esposito et al., 2014). Experiments were terminated for analysis 7 days after rabies injections. All tissue analyzed was cryoprotected in 30% sucrose/PBS and cut on a cryostat (brain: 80 mm sagittal; spinal cord: 40-80 mm transverse sections). Immunohistochemistry and Imaging Antibodies used in this study were: chicken anti-GFP (Invitrogen), guinea pig anti-vGlut1 (Chemicon), goat anti-ChAT (Chemicon), mouse anti-NeuN (Chemicon), mouse anti-Myc (ATCC), mouse anti-V5 (Invitrogen) and rabbit anti-RFP (Rockland). Fluorophorecoupled secondary antibodies were from Jackson or Invitrogen. Floating tissue sections were incubated with antibodies in individual wells and mounted for imaging in sequential order. High-resolution three-dimensional images were acquired with an Olympus confocal microscope (FV1000) or a custom-made dual spinning-disk microscope (Life Imaging Services GmbH, Basel, Switzerland) as described before (Tripodi et al., 2011). Low-resolution images for brainstem reconstructions were acquired using an Axioscan microscope (Zeiss; 5x objective). 3D Digital Reconstructions and Anatomical Quantifications All brainstem images were aligned using the ImageJ Trak EM program with custom-written MATLAB codes. Rabies-marked neurons were identified manually using the Imaris Spot detection module (Bitplane), and color-coded according to their location based on Paxino’s mouse brain atlas (Paxinos and Franklin, 2012). Nomenclature for brainstem structures used in this study was adopted from (Esposito et al., 2014). Briefly, Rt (Parvicellular reticular nucleus, including PCRt, iRt); Mc (Magnocellular Reticular region; given the lack of defined boundaries, we pooled under this name Lateral Paragigantocellular, Gigantocellular Reticular Nucleus ventral and alpha part); Pn (Pontine Reticular Nucleus, including Pn oral and caudal part); Gi (Gigantocellular Reticular Nucleus); Ve (Vestibular Nucleus, including cerebellar, lateral and medial part); SpVe (Spinal Vestibular Nucleus); R (Red nucleus) and M1 (motor cortex). For identification of rabies-marked cell body positions of spinal neurons, images were decomposed to individual channels and planes. They were aligned using custom-developed MATLAB codes. Rabies-marked neurons were assigned coordinates manually and resulting reconstructions were plotted using R (http://www.r-project.org) as previously described (Tripodi et al., 2011). For quantification of midline-crossing axons and digital reconstructions of synaptic terminals, acquired images were decomposed to individual channels and planes in Imaris (Bitplane). Midline-crossing axons were identified manually on transverse spinal cord sections and normalized to the number of axons detected in the white matter on segmentally-matched sections using the Imaris Spot detection module (Bitplane). To translate coordinates for assigned synaptic locations (Imaris Spot detection module) to a digital reconstruction, coordinates were extracted with MATLAB codes implemented in Imaris and ipsi-lesional synaptic terminals were normalized to that of contra-lesional terminals detected on the same section. Synaptic terminals and isoline density plots (Tripodi et al., 2011) were reconstructed using R. SUPPLEMENTAL REFERENCES Dominici, N., Keller, U., Vallery, H., Friedli, L., van den Brand, R., Starkey, M.L., Musienko, P., Riener, R., and Courtine, G. (2012). Versatile robotic interface to evaluate, enable and train locomotion and balance after neuromotor disorders. Nat. Med. 18, 1142–1147. Tripodi, M., Stepien, A.E., and Arber, S. (2011). Motor antagonism exposed by spatial segregation and timing of neurogenesis. Nature 479, 61–66.
Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. S3
R, T, S
T R, T, S R, T
R, T, S
R, T
R, T
R, T, S
1 2 3 4 5
TEMPORAL FEATURES OF GAIT Cycle duration (s) Cycle velocity (s) Stance duration (s) Swing duration (s) Relative stance duration (% of gait cycle duration)
6 7 8 9
INTERLIMB COORDINATION Correlation between hindlimb and ipsi forelimb oscillations Correlation between hindlimb and contra forelimb oscillations Correlation between whole hindlimb oscillations Double stance duration (% of gait cycle duration)
10 11 12 13 14 15 16 17 18 19 20
LIMB ENDPOINT TRAJECTORIES Stride length (cm) Step length (cm) 3D limb endpoint path length (cm) Maximal backward position of foot (cm) Minimum forward position of foot (cm) Step height (cm) Maximal speed during swing (cm/s) Time of maximal velocity during swing (% of gait cycle duration) Acceleration at swing onset (cm/s^2) Endpoint velocity (cm/s) Orientation of the velocity vector at swing onset (deg)
40 41 42 43 44
JOINT ANGLES AND SEGMENTAL OSCILLATIONS Crest oscillations (deg) Thigh oscillations (deg) Backward Shank oscillations (deg) Foot oscillations (deg) Whole limb oscillations (deg)
45 46 47 48 49
Crest oscillations (deg) Thigh oscillations (deg) Shank oscillations (deg) Foot oscillations (deg) Whole limb oscillations (deg)
50 51 52
Hip joint (deg) Knee joint (deg) Ankle joint (deg)
R, T
53 54
Whole limb abduction (deg) Foot abduction (deg)
R, T
55 56 57
Hip joint (deg) Knee joint (deg) Ankle joint (deg)
R, T
58 59
Whole limb adduction (deg) Foot adduction (deg)
60 61 62 63 64
Crest oscillations (deg) Thigh oscillations (deg) Shank oscillations (deg) Foot oscillations (deg) Whole limb oscillations (deg)
R, T
65 66 67
Hip joint (deg) Knee joint (deg) Ankle joint (deg)
R, T
68 69
Whole limb medio-lateral oscillation (deg) Foot abduction/adduction (deg)
70 71 72 73
VELOCITY Whole limb oscillation velocity (deg/s) Hip joint angle oscillation velocity (deg/s) Knee joint angle oscillation velocity (deg/s) Ankle joint angle oscillation velocity (deg/s)
74 75 76 77
Whole limb oscillation velocity (deg/s) Hip joint angle oscillation velocity (deg/s) Knee joint angle oscillation velocity (deg/s) Ankle joint angle oscillation velocity (deg/s)
Maximum angle velocity
78 79 80 81
Whole limb oscillation velocity (deg/s) Hip joint angle oscillation velocity (deg/s) Knee joint angle oscillation velocity (deg/s) Ankle joint angle oscillation velocity (deg/s)
Amplitude of angle velocity (Max - Min)
R, T
R, T, S
R, T
R, T, S
R, T
21 22
DRAG Drag duration (s) Relative drag duration (% of swing phase duration)
23 24 25 26 27
STABILITY Lateral displacement during swing Stance width (cm) Pelvis maximum vertical movement Pelvis minimum vertical movement Pelvis vertical movement amplitude
28 29 30 31 32 33 34 35
Variability of saggital trunk oscillation Variability in velocity of saggital trunk oscillation Variability of vertical mid-point hip oscillation Variability of medio-lateral mid-point hip oscillation Variability of vertical mid-point shoulder oscillation Variability of medio-lateral mid-point shoulder oscillation Variability of medio-lateral shoulder rotations Variability of medio-lateral hip rotations
Trunk and pelvic position and oscillations
36 37 38 39
Forward motion of body center of mass (cm) Lateral motion of body center of mass (cm) Vertical motion of body center of mass (cm) 3D motion of body center of mass (cm)
Displacement of virtual center of mass
R, T
R, T, S
Base of support
R, T
82
PC ANALYSIS PC1 variance resulting from PCA applied on elevation angles of the hindlimb
R, T
83 84 85
FFT DECOMPOSITION Temporal coupling between crest and thigh oscillation Temporal coupling between thigh and shank oscillation Temporal couling between shank and foot oscillation
86 87 88 89 90 91
CROSS CORRELATION BETWEEN SEGMENTS Correlation between pelvis and thigh oscillation Correlation between thigh and shank oscillation Correlation between shank and foot oscillation Correlation between hip and knee oscillation Correlation between knee and ankle oscillation Correlation between ankle and foot oscillation
R, T
92 93 94 95 96 97
RELATIVE COUPLING BETWEEN SEGMENTS Duration between maximal backward positions of crest and thigh (% of gait cycle duration) Duration between maximal forward positions of crest and thigh (% of gait cycle duration) Duration between backward positions of thigh and shank (% of gait cycle duration) Duration between forward positions of the thigh and shank (% of gait cycle duration) Duration between backward positions of shank and foot (% of gait cycle duration) Duration between forward positions of shank and foot (% of gait cycle duration)
L
98 99 100 101
LADDER PERFORMANCE Relative horizontal position of the foot-rung during stance Relative vertical position of the foot-rung during stance Relative horizontal position of the foot (percent distance between rungs) Performance score (1 Accurate; 2 Slip; 3 Miss)
102 103 104 105 106 107 108 109
REPRODUCIBILITY OF ENDPOINT TRAJECTORY Variability of acceleration at swing onset Variability of endpoint velocity Variability of orientation of the velocity vector at swing onset Consistency of limb endpoint trajectory over 10 successive cycles in X plane Consistency of limb endpoint trajectory over 10 successive cycles in Y plane Consistency of limb endpoint trajectory over 10 successive cycles in Z plane Consistency of limb endpoint trajectory over 10 successive cycles in XY planes Consistency of limb endpoint trajectory over 10 successive cycles in XYZ planes
R, T
110 111 112
LEFT-RIGHT COORDINATION Correlation between limb axis angles (in XY) of considered limb and contralateral limb Coordination between stance left-right (% of gait cycle duration) Coordination between swing left-right (% of gait cycle duration)
R
113 114 115 116
COMPARISON TO REFERENCE JOINT ANGLES Correlation between limb oscillation and reference limb oscillation Correlation between hip oscillation reference hip oscillation Correlation between knee oscillation reference knee oscillation Correlation between ankle oscillation reference ankle oscillation
S, T*
117 118 119 120
EMG Relative onset of flexor EMG burst (% of gait cycle duration) Relative end of flexor EMG burst (% of gait cycle duration) Relative onset of extensor EMG burst (% of gait cycle duration) Relative end of extensor EMG burst (% of gait cycle duration)
S, T*
121 122
Flexor EMG burst duration (s) Extensor EMG burst duration (s)
S, T*
123 124 125 126 127 128
Mean amplitude of flexor EMG burst (mV) Integral of flexor EMG burst Root mean square of flexor EMG burst Mean amplitude of extensor EMG burst (mV) Integral of extensor EMG burst Root mean square of extensor EMG burst
S, T*
129
Co-contraction of extensor and flexor muscle
Forward
R, T
Extension
Abduction
Flexion
Adduction
Oscillation amplitudes (Max-Min)
T
R, T
R, T, S
T
R, T, S
R, T, S
R, T
R, T, S
R, T
R, T, S
R, T
Minimum angle velocity
proximal distal
proximal distal proximal distal
proximal
distal
proximal
distal
Timing
Duration
Amplitude
Figure S1. List of Parameters Used for Kinematic Analysis, Related to Figures 1, 2, 3, and 4 Kinematic and EMG parameters were computed offline using custom written MATLAB scripts. Letters in the left column indicate the specific parameters included in PC analysis for each behavioral task. R: runway (103 parameters); T: treadmill (108 parameters); L: ladder (4 parameters); S: swimming (47 parameters); * only animals with EMG implantation, for details see Extended Experimental Procedures.
S4 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
1
High-resolution kinematic recordings
5
Movies S1-S4
2
Extraction of average coordinate per animal along each PC axis Figure 1C, D; Figure 3D Figure 4C
3D reconstruction and offline computation of kinematic parameters
6
Figure 1B, Figure S1
Factor loading extraction for identified PC Figure S3A
3
Principal component analysis on all mice and parameters
7
PCs are computed to maximize the amount of explained variance
4
Extraction of functional clusters for highly correlated parameters (|0.5| > r) Figure S3A,C, Figure S4A
Representation of individual gait patterns in PC space
8
Histogram plots of identified parameters Figure S3A,C, Figure S4A
Figure 1C, D; Figure 3C
Figure S2. Steps of Locomotor Analysis, Related to Figures 1, 2, 3, and 4 Flowchart depicting the multi-step approach applied for behavioral analysis. A detailed description of each step is available in the Extended Experimental Procedures. Each step in the flowchart references a movie or Figure panel where respective steps are shown.
Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. S5
A
Affected
Affected (35%)
(a) Left-right (b) Hip oscillation amplitude coordination (deg) (a.u.) ***
60
0 -0.2
PC1 18%
Unaffected
Unaffected (65%)
(e) Coordination (f) Shank ankle and knee oscillation (a.u.) amplitude (deg)
(e)
(a) 110 8 88 89 87
9 83 84 86 90 92
96 82
93 94 95 97 111112
Coordination
1
80 60
40 20
Oscillation velocity
0
Flexion
-0.4 -0.6 -0.8
**
81 77 76 73 70
79 80 75 74 78
59 58
**
1
(f)
61 65 64 67 63
69 66 62 54 53 52
60 68 36 46 44 42
49 48 47 45 43 41
*
**
0
Stability
(c) Endpoint control
Egr3-/-
Wild-type
Gait timing Joint and endpoint consistency
Factor loading -1 -0.5 0
2
0.5
1
B
5
2
Factor loading
107105116 114 115113
7 cm/s Swing Stance
Limb
23 cm/s
15 cm/s
-1 -0.5 0
0.5
1
Egr3-/-
Wild-type Crest
Egr3-/-
Wild-type
4
(d) 109108106
10 0
0
22 21 19 18 17 16 15 14 12 11 10
3
20
1
(h)
13 20
1
30
23 24 27 29 31
35 28 26 25
0.6 0.8
50 40
36 37 38 39
backward forward
(h) Foot velocity at swing onset (cm/s)
3 (g)
0.9
0
(g) Step width (cm)
40
0=pelvis
-1
0
50 51 52 53 54
(b) Oscillation amplitude
40 20
56 57 55
Extension
(c) Foot position (d) Endpoint during stance reproducibility (cm) (a.u.)
0.5
72 71
7 cm/s
15 cm/s
23 cm/s
1 cm
Foot
2 cm Left limb
80 deg
Extensor
2 mV
Flexor Left stance Right stance
5 mV
2
1
PC1 - Speed
PC2 - Genotype
Coordination
(i) Coordination between shank and foot (a.u.)
39 36 36 68 36 37 32 26
1
8 82 9
n.s.
(j) Stance duration (s) 0.6
Stability (i)
n.s.
36 88 89 36 87 90 96
13 14
0.4
Endpoint control
0.2
39 36 36 68 36 37
3
**
24 103 19 15
0.5
Oscillation amplitude 0 (j) Factor loading 1
(k) Foot position during stance (cm)
Oscillation amplitude (k) Foot positioning
Segmental coupling
2
49 64 62
* ***
36 6 110 36 7
**
Time (s) 0
C
25 31 29 30 38
0 Wild-type
Egr3-/-
27 37 26 28
Trunk stability
-2.5 backward
0=pelvis
2.5 forward
5
Figure S3. Principal Component Analysis of Runway and Treadmill Locomotion, Related to Figure 1 (A) PC analysis was applied on all gait parameters measured during overground locomotion (average of 10-25 gait cycles/hindlimb/mouse, 103 parameters per gait cycle, n = 22 wild-type and n = 19 Egr3 mutants). Parameters correlating with the genotype-specific PC1 (jfactor loadingj > 0.5) and thus specifically altered in Egr3 mutant mice, were regrouped into functional clusters, which we named for clarity. Comparative analysis revealed an overall shift in gait timing, altered intraand interlimb coordination, increased limb and specific joint oscillation amplitudes, postural instability, as well as altered endpoint control in Egr3 mutant mice. All the numbers refer to Figure S1. Histogram plots report mean values for representative variables (a-d). Parameters that were not affected by the lack of muscle spindle feedback circuits, and therefore, did not correlate with the genotype-specific PC1 (jloading factorj < 0.5). Parameters were regrouped into functional clusters. Histogram plots report mean values for representative variables (e-h). (B) Representative color-coded stick decomposition of hindlimb movement during stepping on a treadmill at 3 different speeds. Below, limb endpoint trajectories, velocity vector at swing onset and concurrent EMG activity of an extensor and a flexor muscle are shown. Dark gray bars indicate stance, while empty spaces correspond to swing. (C) PC analysis applied on averaged values of gait parameters (15–30 gait cycles/mouse/speed, 108 parameters per gait cycle, n = 10 for each genotype) measured during stepping on a treadmill at five different speeds (7–23 cm/s, increments of 4 cm/s). Functional clusters associated with the speed-dependent PC1 (16% explained variance) and histogram plots reporting mean values of representative gait parameters modulated with speed increase (i, j). While treadmill speed was the dominant factor correlating with adjustment of gait features in both wild-type and Egr3 mutant mice, the PC2 still segregated genotypic differences (10% explained variance). Functional clusters associated with the genotype-dependent PC2 and histogram plots reporting mean values of a representative gait parameter associated with PC2 (k) are shown. *p < 0.05; **p < 0.01; ***p < 0.001; n.s.: not significant; error bars: SEM; extensor: medial gastrocnemius; flexor: tibialis anterior; a.u.: arbitrary unit.
S6 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
(e) 70 71 74 78 76 80 79
8 112 111 87 82 88
5
Extensive recovery in wild-type
2
Gait timing
(a) Maximal ankle joint extension (deg)
36 26 25 39
22 21
Trunk stability
Drag
Coordination (d)
4
Oscillation velocity
1
Factor loading
14 13 16 18 11 19 12 10 15 109106108 20
Endpoint control (g) (f) 49 48 43 47 46 44 41
0
-1
Forward/backward oscillations
n.s.
100
0
50 0
-0.7
Partial recovery in wild-type and Egr3-/-
** ***
(d) Endpoint acceleration at swing onset (cm/s)
*** n.s.
* ***
1500
75
(e) Amplitude of ankle oscillation (deg)
*** n.s.
*** ***
80
(f, g) Whole limb oscillation (deg)
*** **
*** ***
60
1000
50
40
0
20
0
0
*** **
500
25
** n.s.
100
*** n.s.
***
0.7
150
Intact Acute 1 week 2 weeks 4 weeks Chronic
Partial recovery in wild-type (c) Drag duration (% of swing phase)
n.s.
*** n.s.
***
200
(b) Left - right coordination (a.u.)
n.s. ***
Joint oscillation amplitudes (b) (c)
ild Eg - t y p r3 e -/-
(a)
57 36 52 55 56 64 62 63 61 67
W
A
-100
-50 0=pelvis
backward
***
C
***
Intact Acute
Intact Acute
0.4
0.5 0.2
0
***
0.6
0.4 3.5
0.8
0
Extensor
Flexor Slopes
0
1
0.5
0
*
Slopes 0.5 Burst duration (s)
Burst duration (s)
1
1
0.5
0
in-phase
2 cm
0.2
0
0
out-of phase
0
0 -0.2 -0.4 -0.6
out-of phase
Amplitude of whole limb oscillation (deg)
***
Left - right alternation
Power stroke Return stroke 2 cm
0.6
1
Step height (cm)
Left-right coordination (a.u.)
Stance duration (s)
Wild-type
7
D
Compensation
Unaltered Stride length Coordination knee (cm) and ankle (a.u.)
Egr3-/-
B
50
forward
80 60 40 20 0
Wild-type Intact 2 weeks Acute 4 weeks Chronic 1 week
Figure S4. Detailed Quantification of Functional Recovery after Spinal Cord Injury, Related to Figure 3 (A) Functional clusters correlating with recovery-related PC1 (30% explained variance), and histogram plots reporting mean values for representative variables that recovered extensively (a, b) or partially (c, d) in wild-type mice. Egr3 mutant mice did not show significant improvement for these parameters from acute to chronic phases. (e-g) shows parameters that exhibited a significant recovery from acute to chronic stages in Egr3 mutants after lateral hemisection. (B) Gait analysis of the contralesional hindlimb prior and three days after a hemisection injury revealed that many parameters were unaltered in the contralesional hindlimb in both genotypes. Histogram plots report mean values for representative variables that were unaltered immediately after injury (left), or contributed to compensating decreased weight-bearing capacity of the ipsilesional hindlimb. Parameters were objectively identified using PC analysis (n = 481 steps; 10 – 20 steps/mouse; 108 parameters per gait cycle; n = 9 wild-type and n = 8 Egr3 mutant mice). (C) Correlation between step cycle duration and extensor or flexor burst duration. Regression lines for a representative wild-type mouse are shown for all time points analyzed. Histogram plots report the slopes of regression lines for the ankle extensor or flexor muscle for each time point (n = 3 wild-type; 50-100 step cycles/mouse/time point, total of 726 steps). (D) Stick decomposition of hindlimb movement for a wild-type mouse during swimming at the chronic stage after hemisection (below: limb endpoint trajectories, limb endpoint velocity vectors at power stroke onset). The polar plot shows coordination between left (ipsilesional) and right hindlimb oscillations (black lines: single gait cycle; red arrow: average of all gait cycles). Histogram plots report mean values for kinematic variables. *p < 0.05; **p < 0.01; ***p < 0.001; n.s.: not significant; error bars: SEM; extensor: gastrocnemius medialis; flexor: tibialis anterior; a.u.: arbitrary unit; acute: 3 days postinjury; chronic: 7 weeks postinjury.
Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. S7
Ipsi-injection (=lesional)
R t M c G Sp i Ve Ve Pn
R M 1 R M 1 R M 1
Chronic
R t M c G Sp i Ve Ve Pn
20
0
0
40
40
30
30
10
R t M c G Sp i Ve Ve Pn
20
20
R M 1
40
R t M c SpGi Ve Ve Pn R M 1
Wild-type
60
40
Egr3-/-
R M 1
0
R M 1
0
R t M c SpGi Ve Ve Pn
10
% rabies-positive neurons normalized to brain
R t M c SpGi Ve Ve Pn
Acute
Lumbar
10
60
Hx Rabies
20
R M 1
Thoracic
30
R M 1
Cervical
Contra-injection (=lesional)
20
R t M c G Sp i Ve Ve Pn
Brainstem
Intact
30
% rabies-positive neurons normalized to brain
R t M c G Sp i Ve Ve Pn
Contra
R t M c SpGi Ve Ve Pn
B Ipsi
R t M c G Sp i Ve Ve Pn R M 1
A
20
* *
10 0 R M 1
R t M c G Sp i Ve Ve Pn
R t M c SpGi Ve Ve Pn R M 1
0
Figure S5. Reduced Injury Responses in Brainstem Pathways of Egr3 Mutant Mice, Related to Figure 5 (A) Schematic diagram illustrating retrograde rabies virus injection strategy to label brainstem neurons with descending projections to ipsilesional lumbar spinal cord below injury (yellow territory). (B) Quantification of descending neurons in brain and brainstem of wild-type and Egr3 mutant mice (same animals as in Figure 5). Numbers are displayed as percentage of total neurons detected in the brain, as opposed to subpopulations of individual nuclei shown in Figure 5. For abbreviations, see Experimental Procedures.
S8 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
A
B
Wild-type Acute
Intact
Chronic
Egr3-/Acute
Intact
Chronic
Cervical
Ipsi Contra Cervical
Rabies
Thoracic
Thoracic
Hx
Lumbar
T10 60
-500 *
500
0 500 µm
Ipsi-lesional PNs Contra-lesional PNs
-500 -400
ta ct Ac ut e C hr on ic
0 Ipsi Contra
0 400 µm 23 ±1%
31
Egr3-/-
2 ±0.1% 2 ±1%
8
32
±2%
37 ±6%
63
Intact
500
±3%
10 23
Acute
Chronic
26
39 ±2%
70
Egr3-/intact
±4%
±3%
-400
0
400
68
1
Wt chronic
Wt intact
-500
±7% ±10%
±13%
Wt acute
0
2 ±2% 14 16
±3%
Egr3-/chronic
-500
500
27
±6% ±2%
±2%
±7%
Egr3-/acute
0 400 µm
1 ±2% 1 ±1% 24 11
23
44
±0.1%
0
-400
±2%
±4%
±5%
E
µm
Wild-type
Ipsi-lesional PNs Contra-lesional PNs
0
D
µm
0 500 µm
Correlation of cell distribution
µm 20
Wild-type
Egr3-/-
-500
40
In
% ipsi-lesional rabies positive neurons
C
0.5
0 Wt
Wt
Wt
Egr3-/- Egr3-/-Egr3-/-
Figure S6. Detour Circuit Formation after Spinal Cord Injury Is Reduced in Egr3 Mutant Mice, Related to Figure 6 (A) Schematic diagram illustrating rabies virus injection strategy to retrogradely label descending spinal projection neurons above lesion with projections to ipsilesional territory below lesion (yellow). Neurons with dotted axons are severed by injury, not accessible for retrograde tracing, and dual crossing ipsilesional neurons are highlighted in magenta. (B) Representative 3D reconstructions of supralesional spinal projection neurons in wild-type (left) and Egr3 mutant (right) mice shown at intact, acute and chronic stages in top-down longitudinal view (top) and transverse section (below) view. The lesion position is indicated by filled triangle, the midline by gray line, and ipsilesional neurons by magenta labels. Pie charts at the bottom indicate percentage of neurons within each color-coded spinal quadrant. (C) Quantification of percentage of ipsilesional rabies positive spinal projection neurons above lesion. Note abrupt decrease at acute stages for both genotypes, but reduction of dual-crossing projection neurons in Egr3 mutants compared to wild-type at chronic stages. (D) Representative rabies-marked neurons in spinal segments below lesion and above injection site (T11-L1) in transverse section view. (E) No detectable differences in distribution of rabies-marked neurons among genotypes or lesion conditions. *p < 0.05; error bars: SEM. Acute, 3 days postinjury; chronic, 7 weeks postinjury.
Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc. S9
0
I oj psi ec tio n pr Con oj tr ec a tio n
0
I oj psi ec tio pr Con n oj tr ec a tio n
Ipsi
Wild-type
Contra-lesional Ipsi-lesional PNs PNs µm
D
Cervical PNs
0
5 4 3 2 1 0 5 4 3 2 1 0
0
1
0
G
0 µm
400
Wild-type
500 0
-500 -400
0
400
0
T8/9 T11/13
Egr3-/-
Wild-type
*
µm
3
F
Midline-crossing axons/10 axon tracts
Contra
Ipsi-lesional
Ipsi
Egr3-/-
Egr3-/-
400
Contra-lesional
Thoracic PNs
0
1.5
-400 -400
E
1
pr
Contra
Wild-type
1.5
Egr3-/-
Ipsi-lesional synaptic terminals (%)
35
Contra-lesional PNs
Ipsi-lesional synaptic terminals (%)
35
Wild-type Egr3-/-
Ipsi/contra-lesional axon tracts in WM
70
Ipsi-lesional PNs
Above Hx
70
C
B
Intact thoracic PNs
Below Hx
Intact cervical PNs
pr
Synaptic terminals (%)
A
T8/9 T11/13
60 40 20 0 60 40 20 0
30 20
*
10 0 30
*
20 10 0
Figure S7. Distinct Reconnection Mechanisms for Spinal Projection Neuron Subpopulations, Related to Figure 7 (A) Cervical and thoracic spinal projection neuron analysis of intact wild-type and Egr3 mutant mice (n = 3 for each genotype). (B) Ipsi- and contra-lesional spinal projection neuron descending tract analysis above and below injury. Representative images of tracts and quantification are shown for wild-type and Egr3 mutant mice. Note absence of marked white matter tracts on ipsilesional side below lesion for both genotypes and injections. (C) Low-resolution picture and scheme to illustrate unilaterality of cervical injections. (D) Quantification and distribution of synaptic terminals above lateral hemisection at chronic stages. (E) Low-resolution picture and scheme to illustrate unilaterality of thoracic injections at chronic stages. (F) Frequency analysis of midline-crossing axons at chronic stages originating from ipsi- or contra-lesional thoracic spinal projection neurons, normalized to number of marked axons in contralesional white matter tracts below lesion (ipsilesion: n = 4 each for wild-type and Egr3 mutant, contralesion: n = 5 for wild-type and n = 3 Egr3 mutant). (G) Quantitative analysis of distribution and density of synaptic terminals in the spinal cord below injury, originating from ipsi- or contra-lesional thoracic spinal projection neurons (yellow marks ipsilesional territory below injury; ipsilesion: n = 6 for wild-type and n = 7 for Egr3 mutant, contralesion: n = 7 each for wild-type and Egr3 mutant). Contour plots show overall distribution of terminals, and bar graphs display percentage of ipsilesional synaptic terminals at analyzed segmental spinal levels. *p < 0.05; error bars: SEM; PN: projection neuron.
S10 Cell 159, 1626–1639, December 18, 2014 ª2014 Elsevier Inc.
