Figures and data in Linking spinal circuit reorganization to recovery after thoracic spinal cord injury

Figures
Tables
Additional files

14 figures, 4 tables and 1 additional file

Figures

Figure 1

Download asset Open asset

Model concept.

(A) Organization of the spinal locomotor circuitry (intact). (B, C) Neural structures affected by thoracic hemisection (B) or contusion (C) injuries. Spheres represent neural populations involved in commissural and long propriospinal pathways. Descending drives and synaptic interactions are shown by arrowheads. Decreased color intensity in (C) signifies partial disruption of pathways by contusion injury. CINs, commissural interneurons; LPNs, long propriospinal interneurons.

Figure 2

Download asset Open asset

Detailed model schematic.

Spheres represent neural populations. Excitatory drives and connections are marked by arrowheads; inhibitory connections are marked by circles. RG, rhythm generator; In, interneuron; CINs, commissural interneurons; LPNs, long propriospinal neurons; a-, ascending; d-, descending. See text for details. Adapted from Danner et al., 2017, and Zhang et al., 2022, with modifications.

Figure 3

Download asset Open asset

Gait expression in pre-injury/intact model and rats.

(A1, A2) Extensor/stance phases (upper panels) and instantaneous normalized phase differences (bottom panels) of representative bouts for the model (A1) and a rat (A2). (B1, B2) Average extensor/stance phases for each gait (upper panels; error bars indicate circular standard deviations) and circular plots of average normalized phase differences for each gait (bottom panels; vector length corresponds to mean resultant length, R) expressed in the intact model (B1) and rats (B2). Detailed statistical results for rats are reported in Danner et al., 2023. H.-b., half-bound; RG, rhythm generator.

Figure 4

Download asset Open asset

Frequency-dependent distribution of normalized phase differences, gaits, and phase durations in the model and in rats.

(A1, A2) Scatter plots of normalized phase differences against frequency of locomotor oscillations in the intact model (A1) and rats (A2). Each dot represents one step cycle. Gaits are classified for each step cycle and color-coded. (**B1, B2**) Distribution of gaits vs. locomotor frequency in the model (B1) and rats (B2). Due to the low prevalence of lateral-sequence and canter steps, these gaits were omitted in (B1) and (B2). (C1, C2) Flexor and extensor phase duration in the model (C1) and duration of swing and stance in rats (C2) against frequency of locomotor oscillations. The same number of step cycles is shown for the model and animals; model step cycles were randomly sampled. l-, left; r-, right; -f, fore RG/limb; -h, hind RG/limb; RG, rhythm generator.

Figure 5

Download asset Open asset

Gait prevalences, variability of interlimb coordination, and gait transition probabilities in model and rats.

(A1) Prevalence of gaits in the intact model and following simulated recovery from hemisection and contusion injury. (A2) Prevalence of each gait across intact rats and rats after recovery from hemisection and contusion injury (recalculated from Danner et al., 2023). (**B1, B2**) Means of the deviations from the moving average of each phase difference for intact case and after recovery from hemisection and contusion injuries for the model (B1; error bars denote standard deviations) and rats (B2; error bars denote 95% confidence intervals). Detailed statistical results for rats are reported in Danner et al., 2023. (C1–C3) Matrices of gait transition probabilities in the intact model and following recovery from hemisection and contusion injury. (D1–D3) Gait transition graphs, where nodes represent gaits (size is proportional to their prevalence) and edges represent gait transitions (line widths are proportional to their frequency of occurrence). l-, left; r-, right; -f, fore RG/limb; -h, hind RG/limb; nd, not defined; diag., diagonal; lat., lateral; seq., sequence; RG, rhythm generator.

Figure 6

Download asset Open asset

Conceptual schematic of the impact of hemisection injury on the spinal locomotor circuitry (A) and its reorganization after recovery from hemisection predicted by the model (B).

In (B), affected long propriospinal neuron (LPN) connections recovered functionally through detour pathways. Drives to lumbar and cervical commissural interneurons (CINs) were altered to strengthen left–right alternation. Drive to the ipsilesional lumbar rhythm generator (pink arrow at the bottom) was substituted by regenerated brainstem input and/or afferent feedback.

Figure 7

Download asset Open asset

Gait expression in the model and rats following recovery after hemisection.

(A1, A2) Extensor/stance phases (upper panels) and instantaneous normalized phase differences (bottom panels) of representative bouts for the model (A1) and a rat (A2). (**B1, B2**) Average extensor/stance phases (upper panels; error bars indicate circular standard deviations) and circular plots of average normalized phase differences for each gait (bottom panels; vector length corresponds to the mean resultant length, R) expressed in the post-hemisection model (B1) and rats (B2). Detailed statistical results for rats are reported in Danner et al., 2023. (C1) Prevalences of lead RG in the intact model and following simulated recovery after hemisection for gallop and canter. (C2) Prevalence of leading limbs (left or right forelimb that touches down second) pre-injury (intact) and after recovery of hemisection for gallop and canter in rats. Adapted from Danner et al., 2023. T., trot; C., canter; RG, rhythm generator.

Figure 8

Download asset Open asset

Frequency-dependent distribution of normalized phase differences in the model and in rats following recovery after hemisection.

(A1, A2) Scatter plots of normalized phase differences are plotted against frequency of locomotor oscillations. Each dot represents one period/step cycle. Gaits are classified for each period/step cycle and color-coded. (**B1, B2**) Distribution of gaits vs. locomotor frequency in the post-hemisection model (B1) and rats (B2). The same number of step cycles is shown for the model and animals; model step cycles were randomly sampled. l-, left; r-, right; -f, fore RG/limb; -h, hind RG/limb; RG, rhythm generator.

Figure 9

Download asset Open asset

Bifurcation diagrams of the intact model (A1), following simulated recovery after hemisection (A2), and for model versions where only long propriospinal neuron (LPN) connections were affected (40% of the pre-injury values; A3) or only brainstem drive to the ipsilesional hind RG was reduced (to 90% of the pre-injury value; A4).

Diagrams are plotted against the bifurcation parameter α and with reduced noise, $σ_{N o i s e}$ =5 fA. Normalized phase differences of 0.5 correspond to alternation, whereas phase differences of 0 or 1 correspond to synchronization. (B1–B4) Dependency of locomotor oscillation frequency on parameter α. Blue and red lines indicate stable phase differences or frequency with stepwise increase and decrease of parameter α, respectively. Colored areas indicate the expressed gait. l-, left; r-, right; -f, fore RG/limb; -h, hind RG/limb; RG, rhythm generator.

Figure 10

Download asset Open asset

Conceptual schematic of the contusion injury (A) and following recovery after contusion in the model (B).

The weights of the long propriospinal neuron (LPN) connections between the cervical and lumbar compartments were significantly reduced. Brainstem drive to the lumbar rhythm generators (RGs) was substituted with additional drives to these RGs (pink arrows at the bottom). Brainstem drives to the cervical RG were adjusted to match oscillation frequency of lumbar RGs (gray arrows at the top). Inhibitory drives to cervical V0v commissural interneurons (CINs) were reduced to secure fore left–right alternation (pink arrows at the top). Commissural pathways in the lumbar compartment were reorganized to secure hind left–right alternation (pink arrows at the bottom).

Figure 11

Download asset Open asset

Gait expression in the model and rats following recovery after contusion.

(A1, A2) Extensor/stance phases (upper panels) and instantaneous normalized phase differences (bottom panels) of representative bouts for the model (A1) and a rat (A2). (B1, B2) Average extensor/stance phases for each gait (upper panels; error bars indicate standard deviations) and circular plots of average normalized phase differences for each gait (bottom panels; vector length corresponds to mean resultant length, R) expressed in the post-contusion model (B1) and rats (B2). Detailed statistical results for rats are reported in Danner et al., 2023. D.sq., diag. seq., diagonal-sequence; lat. seq., lateral-sequence; RG, rhythm generator.

Figure 12

Download asset Open asset

Frequency-dependent distribution of normalized phase differences in the model (A1) and in rats (A2) following recovery after contusion.

(A1, A2) Scatter plots of normalized phase differences are plotted against frequency of locomotor oscillations. Each dot represents one period/step cycle. Gaits are classified for each period/step cycle and color-coded. (**B1, B2**) Distribution of gaits vs. locomotor frequency in the model (B1) and rats (B2). l-, left; r-, right; -f, fore RG/limb; -h, hind RG/limb; RG, rhythm generator.

Figure 13 with 3 supplements

Download asset Open asset

Local sensitivity analysis of hemisection (A) and contusion (B) models.

Shown are Sobol first-order (S1) and total-order (ST) sensitivity indices quantifying how small parameter variations (80–125% of baseline) affected locomotor behavior, measured as the Earth Mover’s Distance (EMD) across six interlimb phase differences and locomotor frequency. Parameters include drive to the sublesional lumbar flexor rhythm generator centers (RG-F; ipsilesional in hemisection, bilateral in contusion), connection weights of cervical and lumbar V0 and V3 commissural interneuron (CIN), and long propriospinal neuron (LPN) pathways. Error bars indicate 95% confidence intervals.

Figure 13—figure supplement 1

Download asset Open asset

Effects of parameter variation on model output in the hemisection model.

Shown are scatter plots of normalized phase difference vs. frequency distributions (A1–E1) and corresponding gait–frequency distributions (A2–E2) for the post-injury baseline model (A) and for four parameters varied individually to the lower (80%) and upper (125%) bounds of the sensitivity analysis range: drive to the ipsilesional lumbar flexor rhythm generator center (RG-F) (B), recovered long propriospinal neuron (LPN) connections (C), lumbar V0 commissural interneuron (CIN) connections (D), and lumbar V3 CIN connections (E). All other parameters were held at their baseline values. l-, left; r-, right; -f, fore; -h, hind RG.

Figure 13—figure supplement 2

Download asset Open asset

Effects of parameter variation on model output in the contusion model.

Shown are scatter plots of normalized phase difference vs. frequency distributions (A1–E1; random sample of 500 locomotor cycles) and corresponding gait–frequency distributions (A2–E2) for the post-injury baseline model (A) and for four parameters varied individually to the lower (80%) and upper (125%) bounds of the sensitivity analysis range: drive to the lumbar flexor rhythm generator centers (RG-Fs) (B), ascending long propriospinal neuron (LPN) connections (C), lumbar V0 commissural interneuron (CIN) connections (D), and lumbar V3 CIN connections (E). All other parameters were held at their baseline values. l-, left; r-, right; -f, fore; -h, hind RG; nd, not defined (gait); lat. seq., lateral-sequence; diag. seq., diagonal-sequence.

Figure 13—figure supplement 3

Download asset Open asset

Predicted Earth Mover’s Distance to baseline as a function of lumbar V0/V3 commissural interneuron connectivity.

Contour plots of surrogate-predicted Earth Mover’s Distance as a function of lumbar V0 and V3 commissural interneuron connection strengths (80–125% of baseline; same range as used for the sensitivity analysis) in the hemisection (A) and contusion (B) models, with all other parameters held at their baseline values.

Author response image 1

Download asset Open asset

Variability of phase differences as a function of spared long propriospinal neuron connection weights (hemisection model).

Tables

Table 1

Connection weights in the intact model.

Source	Target ( $w_{i j}$ )
Within cervical and lumbar circuits
RG-F	i-InF (0.4), i-V0_D (0.7), i-V2a (1)
RG-E	i-InE (0.4), i-V3-E (0.35), i-Sh2 (0.5)
IniF	i-RG-E (–1)
IniE	i-RG-F (–0.1)
V2a	i-V0_V (1)
V0_V	c-Ini (0.6)
V0_D	c-RG-F (–0.07)
V3-E	c-RG-E (0.02)
Within cervical circuits
RG-F	i-dLPNi (0.7), i-dV0_D (0.5), i-dV2a (0.5)
Ini	i-RG-F (–0.0375)
dV2a	i-dV0_V (0.9)
Within lumbar circuits
RG-F	i-V3-F (0.4), i-aV3 (0.3)
Ini	i-RG-F (–0.075)
V3-F	c-RG-F (0.03)
V3-E	c-InE1 (1)
InE1	c-RG-E (–0.045)
Between cervical and lumbar circuits
dSh2	ih-RG-F (0.005)
aSh2	if-RG-F (0.04)
dLPNi	ih-RG-F (–0.01)
dV0_D	ch-RG-F (–0.075)
dV0_V	ch-RG-F (0.02)
aV3	cf-RG-F (0.065)

i-, ipsilateral; c-, contralateral; f-, fore; h-, hind.

Table 2

Hemisection: differences to pre-injury model.

Connection weights
Source	Target $(w_{i n t a c t} \to w_{h e m i s e c t i o n})$
hl-aV3	fl-RG-F (0.065 → 0.026)
hr-aSh2	fr-RG-F (0.04 → 0.016)
fr-dSh2	hr-RG-F (0.005 → 0.002)
fr-dLPNi	hr-RG-F (–0.01 → –0.004)
fl-dV0_D	hl-RG-F (–0.075 → –0.03)
fl-dV0_V	hl-RG-F (0.02 → 0.008)
Drive parameters
Target	$d_{{E, I}, i n t a c t} \to d_{{E, I}, h e m i s e c t i o n}$
rh-RG-F	$d_{E} = 0.1 \to d_{E} = 0.009$
h-V0_V	$d_{E} = 0.15 \to d_{I} = 0.075$
f-V0_V	$d_{I} = 0.25 \to d_{I} = 0.125$

f-, fore; h-, hind; l-, left; r-, right.

Table 3

Contusion: differences to pre-injury model.

Connection weights
Source	Target $(w_{i n t a c t} \to w_{c o n t u s i o n})$
h-aV3	cf-RG-F (0.065 → 0.00325)
h-aSh2	if-RG-F (0.04 → 0.002)
f-dSh2	ih-RG-F (0.005 → 0.00025)
f-dLPNi	ih-RG-F (–0.01 → –0.0005)
f-dV0_D	ch-RG-F (–0.075 → –0.00375)
f-dV0_V	ch-RG-F (0.02 → 0.001)
Drive parameters
Target	$d / b_{{E, I}, i n t a c t}$ → $d / b_{{E, I}, c o n t u s i o n}$
h-V0_D	$d_{I} = 0.15 \to 1. d_{I} = 0.0; b_{I} = 0.0 \to b_{I} = 0.2$
h-V0_V	$d_{I} = 0.25 \to 1. d_{I} = 0.0; b_{I} = 0.0 \to b_{I} = 0.03$

i-, ipsilateral; c-, contralateral; f-, fore; h-, hind.

Table 4

Idealized gaits.

	Normalized phase differences			Gait
	LR	HL	diag.
One-beat	0	0	0	Pronk
Two-beat	1/2	1/2	0	Trot
	0	1/2	1/2	Bound
	0	2/3	2/3	Bound
	1/2	0	1/2	Pace
Three-beat	0	1/3	2/3	Half-bound
	0	2/3	1/3	Half-bound
	2/3	1/3	0	Canter
	1/3	1/3	2/3	Canter
	1/3	2/3	0	Other
	2/3	2/3	1/3	Other
	1/3	2/3	2/3	Other
	2/3	1/3	1/3	Other
	1/3	0	2/3	Other
	2/3	0	1/3	Other
	1/3	2/3	1/3	Other
	2/3	1/3	2/3	Other
Four-beat	3/4	1/4	2/4	Rotary gallop
	1/4	3/4	2/4	Rotary gallop
	3/4	2/4	1/4	Transverse gallop
	1/4	2/4	3/4	Transverse gallop
	2/4	1/4	3/4	Lateral-sequence
	2/4	3/4	1/4	Diagonal-sequence

LR: hind left–right normalized phase difference, HL: homolateral normalized phase difference; diag.: diagonal normalized phase difference. Adapted from Danner et al., 2023.

Additional files

MDAR checklist: https://cdn.elifesciences.org/articles/107480/elife-107480-mdarchecklist1-v1.docx
Download elife-107480-mdarchecklist1-v1.docx

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Mendeley

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

Natalia A Shevtsova
Andrew B Lockhart
Ilya A Rybak
David SK Magnuson
Simon M Danner

(2026)

Linking spinal circuit reorganization to recovery after thoracic spinal cord injury

eLife 14:RP107480.

https://doi.org/10.7554/eLife.107480.3

Share this article

Cite this article

Model concept.

Detailed model schematic.

Gait expression in pre-injury/intact model and rats.

Frequency-dependent distribution of normalized phase differences, gaits, and phase durations in the model and in rats.

Gait prevalences, variability of interlimb coordination, and gait transition probabilities in model and rats.

Conceptual schematic of the impact of hemisection injury on the spinal locomotor circuitry (A) and its reorganization after recovery from hemisection predicted by the model (B).

Gait expression in the model and rats following recovery after hemisection.

Frequency-dependent distribution of normalized phase differences in the model and in rats following recovery after hemisection.

Conceptual schematic of the contusion injury (A) and following recovery after contusion in the model (B).

Gait expression in the model and rats following recovery after contusion.

Frequency-dependent distribution of normalized phase differences in the model (A1) and in rats (A2) following recovery after contusion.

Local sensitivity analysis of hemisection (A) and contusion (B) models.

Effects of parameter variation on model output in the hemisection model.

Effects of parameter variation on model output in the contusion model.

Predicted Earth Mover’s Distance to baseline as a function of lumbar V0/V3 commissural interneuron connectivity.

Variability of phase differences as a function of spared long propriospinal neuron connection weights (hemisection model).

Connection weights in the intact model.

Hemisection: differences to pre-injury model.

Contusion: differences to pre-injury model.

Idealized gaits.

MDAR checklist

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)