Long ascending propriospinal neurons are heterogenous and subject to spinal cord injury induced anatomic plasticity

Abstract

Long ascending propriospinal neurons (LAPNs) are a subset of spinal interneurons that provide direct connectivity between distant spinal segments. Here, we focus specifically on an anatomically defined population of “inter-enlargement” LAPNs with cell bodies at L2/3 and terminals at C5/6. Previous studies showed that silencing LAPNs in awake and freely moving animals disrupted interlimb coordination of the hindlimbs, forelimbs, and heterolateral limb pairs. Surprisingly, despite a proportion of LAPNs being anatomically intact post- spinal cord injury (SCI), silencing them improved locomotor function but only influenced coordination of the hindlimb pair. Given the functional significance of LAPNs pre- and post-SCI, we characterized their anatomy and SCI-induced anatomical plasticity. This detailed anatomical characterization revealed three morphologically distinct subsets of LAPNs that differ in soma size, neurite complexity and/or neurite orientation. Following a mild thoracic contusive SCI there was a marked shift in neurite orientation in two of the LAPN subsets to a more dorsoventral orientation, and collateral densities decreased in the cervical enlargement but increased just caudal to the injury epicenter. These post-SCI anatomical changes potentially reflect maladaptive plasticity and an effort to establish new functional inputs from sensory afferents that sprout post-SCI to achieve circuitry homeostasis.

Introduction

By definition, propriospinal neurons (PNs) have their somata and axon terminals wholly contained within the spinal cord. Broadly, PNs can be divided into subpopulations of “short” and “long”, the former having axonal projection(s) spanning 3 or fewer spinal segments and the latter projecting 4 or more spinal segments (Leibinger et al., 2021). PNs can further be divided into subpopulations based on embryonic lineage, neurotransmitter phenotype, electrophysiologic characteristics, connectivity, morphology or anatomy (Zholudeva & Lane, 2022). Here, we have focused on a subset of long ascending propriospinal neurons (LAPNs) that are anatomically defined as inter-enlargement, having somata at lumbar level 2/3 (L2/3) and axon terminals at cervical level 5/6 (C5/6).

The majority of these LAPN somata are found in the intermediate gray matter of the L2/3 segments, have axons which ascend in the outer rim of the white matter via the ventrolateral funiculi (VLF), and project to C5/6. Roughly 50% of LAPNs project ipsilaterally while 50% have axons that cross the midline at or near the level of the cell body and ascend in the contralateral VLF (Brown et al., 2021; Pocratsky et al., 2017; Reed, Shum-Siu, Onifer, & Magnuson, 2006).

Independent of motoneuron loss, kainic acid injections into the intermediate gray matter of L2, where LAPN somata reside, resulted in severe locomotor impairments, indicating that neurons in the intermediate gray matter of L2 are important for locomotor function (Hadi, Zhang, Burke, Shields, & Magnuson, 2000; Magnuson et al., 1999). More recently we used a dual-viral system to conditionally and reversibly silence LAPNs in awake and freely moving rats and found that LAPNs are involved in maintaining interlimb coordination of the limb pairs at each girdle (forelimb-forelimb and hindlimb-hindlimb). Silencing these neurons resulted in a partial decoupling of the forelimb and hindlimb pairs and a secondary disruption of heterolateral forelimb-hindlimb coordination (Pocratsky et al., 2020; Shepard et al., 2021). Interestingly, the behavioral phenotype induced by silencing LAPNs was restricted to non-exploratory overground locomotion on a high friction surface, indicating that LAPNs are a flexible, task-specific network of neurons involved in securing interlimb coordination in a context-specific manner.

SCI bidirectionally disrupts the propagation of signals past the injury epicenter leading to functional deficits. However, even if SCI patients are classified as neurologically complete, most injuries are anatomically incomplete (Kakulas, 2004). The spared tissue may provide a means of relaying top-down or bottom-up signals to targets caudal or rostral to the injury, respectively. As LAPN axons ascend in the most superficial layers of the VLF (the outer rim of white matter), where tissue sparing is most likely to occur, they are anatomically situated to be involved in injury-induced plasticity. Bareyre et al. and Filli et al. showed that long descending propriospinal neurons with axons projecting caudal to the injury receive higher numbers of inputs from supraspinal motor centers after an injury, effectively forming detour circuits post-SCI (Bareyre et al., 2004; Filli et al., 2014). This post-SCI rewiring is believed to be an effort to relay top-down signals to spinal levels caudal to the injury. As LAPN axons ascend superficially in the VLF where tissue sparing is most likely to occur, we hypothesized that they would also exhibit injury-induced anatomical plasticity.

Surprisingly, Shepard et al. showed that silencing LAPNs at 6 weeks post-SCI resulted in improved locomotor function (Shepard et al., 2021). These improvements were characterized by an increase in the number of plantar steps, improved hindlimb-hindlimb coordination, and normalized spatiotemporal gait indices. These improvements were not only statistically significant but also functionally meaningful with visibly improved hindlimb extension, tail position, and trunk stability. These data suggest that post-SCI, LAPNs somehow play a detrimental role related to locomotor recovery/function and are likely involved in maladaptive injury-induced plasticity that is alleviated by silencing. These maladaptive changes may result from altered inputs to LAPNs and/or changes in the areas/neurons innervated by the LAPNs. However, as English et al. is the only study to anatomically classify LAPNs based on morphology and did so qualitatively in cats, our goal was to anatomically and morphologically characterize LAPNs in rats before and after SCI to evaluate SCI-induced neuroplasticity (English, Tigges, & Lennard, 1985).

Methods

All animal procedures were reviewed and approved by the University of Louisville Institutional Animal Care and Use Committee and Institutional Biosafety Committee. A total of n=8 female Sprague Dawley rats (200–220 g: Inotiv Maryland Heights, MO) were divided into two groups (N=4 per group). Prior to surgical procedures animals were housed two per cage with ad libitum food and water, and a standard 12:12 light/dark cycle. Animals were handled by researchers daily for several days prior to any testing.

Stereotaxic injections

Glass micropipettes for stereotaxic spinal cord injections were pulled from borosilicate glass capillaries (World Precision Instruments, Inc., Sarasota, FL) using a micropipette puller (Sutter Instrument Co., Novato, CA), were trimmed to an external diameter of 25 μm, beveled using a micropipette beveller (World Precision Instruments, Inc., Sarasota, FL), and sterilized with 100% ethanol prior to use. The morning of injection surgeries, individual viruses were loaded into pipettes and were kept on ice between surgeries to minimize viral degradation.

Animals were anesthetized with a mixture of ketamine, xylazine, and acepromazine (40, 2.5, and 1 mg/kg, i.p.), and supplemented with 1–2.5% isoflurane in 98% oxygen at a rate of 1 L/min as needed. Prior to surgical openings, animals were given 5 mL of saline (s.c.), buprenorphine (10 mg/kg, s.c.), and meloxicam (1.5 mg/kg, s.c.). A laminectomy and durotomy were performed at thoracic vertebrae 12 to expose the L2 & L3 spinal segments. Animals were then placed in a custom-built spinal stabilization unit to stabilize the thoracic spine, and a laminectomy and durotomy were performed at cervical vertebrae 5 and 6 to expose the C5 and C6 spinal segments (Zhang et al., 2008). Using a stereotaxic device (World Precision Instruments, Inc.) HiRet-Lenti-Cre (produced by Boston Children’s, titer = 1.18 x 10¹² gc/mL) was injected bilaterally into the intermediate gray matter of C5/6 (0.55 mm mediolateral, 1.2 mm dorsoventral, 1.3 mm rostrocaudal). Two injection sites on each side of the spinal cord were targeted; at each site two, 0.25 μL boluses were injected with 2 minutes between to allow viruses to spread through the tissue, mitigate extravasation from the injection site and minimize pressure exerted on the tissue (Figure 1A, only unilateral injections shown for clarity). Following cervical injections, the spinal stabilizers were moved to T12 for lumbar injections. AAV2-CAG-FLEx-GFP (produced by the UNC Vector Core, titer = 4 x 10¹² gc/mL) was injected bilaterally into the intermediate gray matter of L2/3 (0.5 mm mediolateral, 1.35 mm dorsoventral, 1.3 mm rostrocaudal) following the same procedures as for the cervical injections (Figure 1A, only unilateral injections shown). Incision sites were sutured in layers and wounds closed with surgical staples. Post-surgery, animals were single housed for 7 days, administered saline 2x/day for 3-5 days, buprenorphine 3x/day for 3 days, and meloxicam 1x/day for 3 days at previously mentioned dosages.

Figure 1.

Experimental design and overview for labeling and reconstruction of long ascending propriospinal neurons (LAPNs) A. HiRet-Lenti-Cre injected bilaterally at cervical 5-6 and AAV2-FLEx-GFP bilaterally at lumbar 2-3 (only ipsilateral injections shown for illustration) to label both ipsilaterally and contralaterally projecting LAPNs B. Maximum intensity projection of LAPN cell bodies and neurites at L2-3, following viral labeling and iDISCO tissue clearing. White lines depict edges of gray matter. C. Volume rendering of L2-3 confocal image depth coded to visualize rostrocaudal location of viral labeled LAPN somata and neurites. D. D’ traced LAPN (red) overlaid on maximum intensity projection D’’-D’’’’ Visualization of traced neuron (yellow) in multiple planes with convex hull (blue).

Behavioral assessments

Three weeks following intraspinal injections, animals underwent pre-SCI baseline testing. To assess gross hindlimb locomotor function the Basso, Beattie, Bresnahan open-field locomotor scale (BBB) was used by trained assessors blinded to the experimental groups (Basso, Beattie, & Bresnahan, 2002). Overground gait analyses and interlimb phase calculations were performed as previously described (Pocratsky et al., 2017; Pocratsky et al., 2020; Shepard et al., 2021).

Phase values from overground recordings were converted to a linear scale to eliminate any lead limb preference differences and to allow for linear plotting.

Thermal nociceptive thresholds were assessed using a radiant heat tail flick device (Grau, 1984). Animals were briefly wrapped in a small towel with the tail exposed and placed on the device. Once the tail was flat and immobile, the heat source was started, tail movement exceeding 0.5 cm terminated the heat source, and the latency was recorded. A minimum of 2 minutes separated trials. Thermal nociceptive thresholds of the hind paws were tested using the Hargreaves device, where animals were placed on a warmed glass surface (32°C) and allowed to acclimate to the environment for 10-15 minutes. An infrared light was shone onto the plantar surface of the foot and the latency to movement of or attending to the paw was recorded. At least 2 minutes elapsed between trials.

Spinal cord injury surgeries

Animals were anesthetized with the same anesthetic regime used for intraspinal injections and given the same pre-operative drugs. A T9 laminectomy was performed to expose the T10 spinal segment, the dura retained intact, and the T8/T10 vertebral segments were stabilized in custom-built spine stabilizers. The Infinite Horizons Impactor (PSI, Fairfax Station, VA) was centered in the laminectomy window and the spinal cord impacted with a 150 kdyn target force (Scheff & Roberts, 2009). Sham animals underwent the same procedures minus the SCI. Surgical closure and post-operative care were the same as following intraspinal injections, with the addition of manual bladder expression as needed. One-week post-SCI the BBB open field locomotor assessment was performed to functionally confirm the injuries. All the previously described behavioral assessments were performed again at 5-weeks post-SCI.

Tissue processing and microscopy

Following the week 5 behavioral assessments, animals were anesthetized using a cocktail of ketamine, xylazine, and acepromazine, (40, 2.5, and 1 mg/kg, i.p.) and transcardially perfused with chilled phosphate-buffered saline (PBS, pH 7.4) followed by chilled 4% paraformaldehyde (PFA). Spinal cords and brains were harvested and post-fixed in 4% PFA on ice for 1-2 hours (spinal cords) or 3-4 hours (brains), and all tissues were transferred to 30% sucrose for 3-4 days at 4°C. The C5-6 and T9-T11 spinal segments and brainstems were isolated, embedded in tissue freezing medium, cryosectioned at 30 μm, slide mounted in sets, and stored at −20°C.

To evaluate spared white matter, injury epicenters (T10) were stained using FluoroMyelin (Invitrogen, F34652, Waltham, MA) following the manufactures protocol, at a dilution of 1:150, for 60 minutes at room temperature and coverslipped with Fluoromount G (Southern Biotech, 0100-01, Birmingham, AB).

For assessment of LAPN axons/sprouts, GFP from virally-labeled LAPN axons was enhanced using immunohistochemistry in thoracic spinal cord sections immediately rostral (T9) and caudal (T12) to the injury epicenter. Sections were warmed at 37°C for 30 min, rehydrated in room temperature PBS for 10 min, incubated in a blocking solution made of nine parts milk solution (bovine serum albumin [BSA]), 0.75 g of powdered skim milk, and 14.25 ml of 0.1% PBS with Tween 20 (PBST) and one part 10% normal donkey serum (NDS) for 60 minutes. Blocking was followed by a 10 min wash in PBS and an overnight incubation at 4°C with rabbit anti-GFP at 1:500 (Abcam, ab290, Waltham, MA). Tissue was then washed 3 times alternating between PBS and PBST for 10 minutes each, followed by secondary antibody (Donkey anti-rabbit Plus 488, Thermofisher, A32790, Waltham, MA) incubation at 1:200 at room temperature for 1 hour. Tissue sections were washed 3 more times alternating between PBS and PBST for 10 minutes each and coverslipped with Fluoromount-G (Southern Biotech, 0100-01, Birmingham, AB). Brainstem sections were stained in the same manner but prior to coverslipping were incubated at room temperature with fluorescent Nissl (NeuroTrace 640/660, ThermoFisher, N21483, Waltham, MA) in PBS at 1:100 for 1 hr and washed 3 times with PBS for 10 minutes each.

Spinal levels L2-3 (where LAPN somata reside) were cleared for large volume imaging. Prior to tissue clearing, the rostral half of the lumbar enlargement was cut coronally into 1.5 mm segments using a spinal cord matrix (Alto Scientific, Eatonton, GA). iDISCO (updated protocol December 2016, https://idisco.info/idisco-protocol/update-history/) was used to clear individual segments (Renier et al., 2014). Segments were incubated in the primary antibody (Chicken anti-GFP, Aves, GFP-1020, Davis, CA) and secondary antibody (Goat anti-chicken 488+, Invitrogen, A32931) solutions for 3 days each and mounted in custom built silicone slide chambers. Solvent based clearing, such as iDISCO, cause neural tissue to shrink approximately 20% resulting in the initial 1.5mm spinal cord segments to be 1-1.2 mm at the time of imaging (Renier et al., 2014).

Imaging

Widefield fluorescent images of thoracic spinal cord and brainstem sections were acquired using a Nikon (Melville, NY) Ti2E inverted microscope with SOLA SE LED white light engine, Hamamatsu Orca Fusion Gen III camera, and DAPI, GFP, and Cy3 filters. For thoracic spinal cords, one slide from each set was imaged, and stitched images were acquired using a CFI60 Plan Apo λ 10X NA lens, with the appropriate filter(s). Injured thoracic tissue was imaged with the same equipment using a CFI Plan Fluor 4X NA lens.

Cleared lumbar segments were imaged with a Nikon C2+ confocal microscope with intelligent acquisition, LUN4 solid state laser launch (405, 488, 561, and 640 nm), DUVb high-sensitivity GaAsP detectors, and an ORCA-Fusion Gen-III sCMOS monochrome camera. Large volume images were acquired with a custom-built JOBS program in NIS-Elements AR (Nikon, version 5.30.05) using CFI60 Plan Apo λ 10X NA lens with the 488 laser. Post-acquisition, volumes were stitched in Nikon Elements (Figure 1B-C).

Microscopy quantification

Spared white matter was calculated for sections spanning 1.5 mm rostrally and 1.5 mm caudally from the injury epicenter to confirm injury severity. The total cross-sectional area of the spinal cord and the lesion boundary were quantified and analyzed. The epicenter of each injury was determined based on the section with the smallest area of spared white matter.

For quantification of LAPN axon collaterals in thoracic and cervical spinal cords heatmaps were generated from 15 sections from each animal, covering 1.7 mm rostocaudally. Heatmaps were generated and quantified using a custom built MatLab (Mathworks, Natick, MA) program to determine the percentage of positive pixels in the spinal gray matter.

To analyze LAPN axon collaterals and sprouting in the brainstem, images were imported into ImageJ and the appropriate anatomic map (Rat Brain Atlas) overlaid. Relevant brainstem nuclei were selected as regions of interest and the number of positive pixels and percent positive area quantified for each nucleus. Tissue from 2 animals (one from each group) were not analyzed due to tissue artifact.

For soma and neurite reconstructions of viral labeled LAPNs, the Simple Neurite Tracer (SNT) plugin in ImageJ was used. Cells were reconstructed using a single point for the soma and neurites manually traced for each cell (Figure 1D). After tracing, the following neurite measures were exported; 1) neurite length which gives a cumulative measure of the length of all traced neurites, 2) convex hull which is the minimum 3-dimensional geometric space needed to contain all neurites, 3) z-projection depth which measures the rostrocaudal distance the cell’s neurites span, and 4) complexity index which provides a calculated measure of branching complexity (Marshak, Nikolakopoulou, Dirks, Martens, & Cohen-Cory, 2007). Soma volumes were measured using the “Fill Manager” in SNT and neurite/trace proprieties were exported to a CSV file. CSV files were analyzed using a custom built MatLab program to quantify neurite directionally of each cell.

Statistical analyses

Agglomerative hierarchical clustering of LAPNs and production of parallel plots were performed in R (Posit Software, version 2023.03.1). Input variables were normalized using min-max scaling, Ward’s clustering method was used, and the number of clusters was determined using the elbow method. Comparison of LAPN cluster morphologic characteristics were compared using one-way ANOVAs followed by Tukey’s HSD post hoc where appropriate in Prism 9 (GraphPad Software, Version 9.5.1). The neurite directionality of LAPN clusters both pre- and post-SCI were analyzed using the Kruskal-Wallis test followed by Dunn’s test where appropriate (performed in Prism 9).

BBB scores were compared using a two-way ANOVA followed by Bonferroni’s multiple comparisons test where appropriate in Prism 9. Tail flick and Hargreave’s test outcomes were compared using unpaired t-tests in Prism 9. Binomial Proportion Test was used to evaluate differences in the proportion of abnormal phase values, defined as greater than 2 standard deviations from baseline measures and regression analyses were performed in Prism 9 for spatiotemporal gait indices.

For comparison of morphologic characteristics of sham versus SCI LAPN clusters, unpaired t-tests were performed in Prism 9. Kruskal-Wallis test followed by Dunn’s test where appropriate were used to analyze axon collateral densities at thoracic spinal levels, and for cervical and brainstem collateral comparisons unpaired - tests used all performed in Prism 9. P values for all analysis were considered statistically significant when p ≤ 0.05, and two-tailed p values are reported for post hoc t-tests.

Results

LAPNs differ based on soma and dendrite morphology

As described above, robust dual-viral labeling of LAPNs allowed for tissue clearing, large volume imaging (2mm of depth), and reliable reconstruction of single LAPNs (Figure 1A-D). Hierarchical cluster analysis based on 11 morphologic characteristics from 572 reconstructed LAPNs revealed that LAPNs group into three distinct clusters (Figure 2A). The resultant clusters contained 171, 198, and 203 cells, respectively. LAPNs in cluster 1 (Figure 2B, B’) are larger, more complex neurons than those in clusters 2 (Figure 2C, C’) and 3 (Figure 2D, D’). This is evident as soma volume (Figure 2E), total neurite length (Figure 2F), convex hull volume (Figure 2G), z-projection depth (Figure 2H), and complexity index (Figure 2I), that are all significantly greater for cluster 1 compared to clusters 2 and 3. Overall size and complexity of LAPNs in clusters 2 and 3 did not differ, however neurite directionality did (Figure 2A, J). Cluster 2 LAPNs have the greatest percentage of neurites that orient mediolaterally and cluster 3 the has the lowest percentage. However, cluster 3 LAPNs have the highest percentage of neurites oriented ventrally and a higher percentage oriented dorsally than cluster 2 (Figure 2J). These resultant clusters reflect three distinct subsets of LAPNs based on soma and neurite morphologies.

Figure 2.

LAPNs differ anatomically based on soma and neurite characteristics. A. Parallel plot with confidence intervals show LAPN clusters from hierarchical cluster analysis based on morphologic characteristics. Dark lines indicate medians for each cluster and lighter bands represent confidence intervals. Y-axis shows scaled units (0-1), so various parameters can be viewed an analyzed together. B, C, D. 3-dimensional rendering of representative LAPNs from each cluster (not shown to scale, for ease of visualization). B’, C’, D’. Representative traced LAPNs (red) overlaid on maximum intensity projections of L2-3. E-I. LAPN clusters differ morphologically (one-way ANOVAs, p<.0001) based on neurite length (Tukey’s HSD post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 1 vs. cluster 3, p<.0001), z-projection depth (Tukey’s HSD post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 1 vs. cluster 3, p<.0001), convex hull (Tukey’s HSD post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 1 vs. cluster 3, p<.0001), soma volume (Tukey’s HSD post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 1 vs. cluster 3, p<.0001), and neurite complexity (Tukey’s HSD post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 1 vs. cluster 3, p<.0001). J. LAPN clusters differ (Kruskal-Wallis tests, p <.0001) based on dorsal (Dunn’s test post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 2 vs. cluster 3, p<.0001), ventral (Dunn’s test post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 2 vs. cluster 3, p<.0001), medial (Dunn’s test post hoc, cluster 1 vs. cluster 2, p<.0001; cluster 1 vs. cluster 3, p<.0001; cluster 2 vs. cluster 3, p<.0001), and lateral (Dunn’s test post hoc, cluster 1 vs. cluster 2, p<.005; cluster 1 vs. cluster 3, p<.0001; cluster 2 vs. cluster 3, p<.05) neurite directionality. E-J. Large center lines show group means and error bars are standard deviations. *p<.05, **p<.01, ***p<.001, **** p<.0001.

Mild SCI causes minimal functional deficits

Mild T10 contusive SCI (150 kDyn) was confirmed in each animal by quantifying the percentage of spared white matter (Figure 3A, B). In addition, assessment of gross locomotor function using the BBB scale confirmed functional deficits at 1-week post-SCI compared to baseline and week 1 scores for sham animals. However, by week 5 BBB scores were similar between the sham and SCI groups (Figure 3C). Sensory testing for thermal hypersensitivity via tail flick response and Hargreaves’ test showed no change from baseline in either group at week 5, which is not surprising given that there were no differences in the gross locomotor function (BBB scores) at week 5 indicating that the delivered injuries were very mild.

Figure 3.

150kdyn T10 SCI produces no gross functional deficits at 5 weeks post-SCI A. Spared white matter by animal following 150kdyn contusive SCI. B. Representative injury epicenter (T9) cross section stained with Fluoromyelin (red). C. Basso, Beattie, Bresnahan locomotor rating scale (BBB) scores for sham and SCI groups. BBB scores decrease (two-way ANOVA, interaction, p=.0001; column factor, p=.0021; group factor, p= .011) at 1-week post-SCI (Bonferroni’s multiple comparisons test, p=.04, sham week 1 20.5 ± 1.0 vs. SCI week 1 13.25 ± 3.2; p<.0001, SCI baseline 21.0 ± 0.0 vs. SCI week 1 13.25 ± 3.2; p <.0001, SCI week 1 13.25 ± 3.2 vs. SCI week 5 19.75 ± 0.96) but recover by week 5. D. Tail flick response and E. Hargreaves thermal sensitivity test show no differences between sham and SCI groups. C-D. Circles represent group means error bars standard deviations. *p<.05 for group interaction, ^†p<.05 for column interaction.

To thoroughly examine locomotor performance, overground gait and kinematics were assessed. Coupling patterns of each limb pair were determined by dividing the initial contact time of one limb by the stride time of the other. Resulting phase values of 0 or 1 indicate synchrony and values of 0.5 indicate alternation of the limb pair. To eliminate discrepancies induced by lead-limb differences and for ease of visualization, phase values were converted to a linear scale of 0.0 - 0.5 or 0.5 – 1.0. Mean phase values of the limb pairs were calculated at baseline and any value >2 standard deviations from these means were considered ‘irregular’, as indicated by the blue boxes (Figure 4A-D). SCI had no impact on left-right alternation of the hindlimb (Figure 4A) or forelimb pairs (Figure 4B). However, at 5 weeks post-SCI homolateral and heterolateral phase values were altered, with homolateral limb pairs showing more synchrony (Figure 4C) and heterolateral limb pairs showing more alternation (Figure 4D). No difference was seen between groups for speed-dependent gait indices of swing time (Figure 4E, J, O), stance time (Figure 4F, K, P), stride frequency (Figure 4G, L, Q), stride time (Figure 4H, M, R), or stride length (Figure 4I, N, S). Together, these data suggest that mild contusive T10 SCI partially decouples the homolateral and heterolateral limb pairs during overground locomotion.

Figure 4.

5 weeks post-SCI limb coupling is impaired but other key features of locomotion are unaltered. A-D. Transformed phase values for limb pairs. There is no disruption in the forelimb (A) or hindlimb (B) pairs, but the SCI group shows disrupted phase values for ipsilateral (C) (binomial proportion test, baseline, n = 12/262 vs. SCI, n = 42/191, p<.001; Sham, n = 16/253 vs. SCI, n = 42/191, p<.001) and contralateral (D) (binomial proportion test, baseline, n = 13/259 vs. SCI, n = 41/178, p<.001; Sham, n = 17/240 vs. SCI, n = 41/178, p<.001) limb pairs. E-S. Relationships between swing time, stance time, stride time, and stride distance are plotted against speed. An exponential decay line of best fit is displayed for stance time (baseline R² =0.749 vs. sham R² =0.824 vs. SCI R²=0.733) and stride time (baseline R² =0.826 vs. sham R² =0.535 vs. SCI R²=0.650), and a linear line of best fit shown for stride frequency (baseline R² =0.801 vs. sham R² =0.496 vs. SCI R²=0.682) and stride length (baseline R² =0.626 vs. sham R² =0.662 vs. SCI R²=0.832). Dotted lines indicate line of best fit and solid lines indicate 95% prediction intervals. ***p<.001

Post-SCI LAPNs resemble uninjured LAPNs but differ in neurite directionality

Hierarchical cluster analysis using 11 morphologic characteristics from 535 reconstructed LAPNs revealed that post-SCI LAPNs can be divided into three distinct clusters (Figure 5A). Based on parallel plots, the pattern of clustering is similar between LAPN clusters in sham and SCI animals (Figure 2A & Figure 4A), and the total number of labeled cells and the number of cells within each cluster did not differ between groups (Supplementary Figure 1A, B). Cluster 1 consisted of larger more complex cells, and clusters 2 and 3 consisted of smaller, less complex LAPNs that differ in neurite directionality. Comparing morphometrics of sham and SCI clusters showed that the clusters are similar based on neurite length (Figure 5B), convex hull (Figure 5C), and complexity (Figure 5E). The differences between sham and SCI were for cluster 2, where after SCI LAPNs had smaller soma volumes (Figure 5C) and greater z-projection depths (Figure 5F).

Figure 5.

(LAPN clusters post-SCI) 5 weeks post-SCI LAPNs morphologically group into 3 clusters, and post-SCI LAPNs morphologically differs from uninjured LAPNs. A. Parallel plot with confidence intervals shows that post-SCI LAPN clusters from hierarchical cluster analysis based on morphologic characteristics. Dark lines indicate medians for each cluster and lighter bands represent confidence intervals. B-F. Uninjured and post-SCI LAPN clusters are similar based on neurite length (B), convex hull (C), and complexity (E) but post-SCI cluster 2 LAPNs differ from uninjured LAPNs in soma volume (unpaired t test, p<.0001, sham 13294 ± 4095 vs. SCI 10931 ± 4880) and z-projection depth (unpaired t test, p=.002, sham 179 ± 78.0 vs. SCI 206.8 ± 109.7). G-I. Post-SCI LAPN neurite directionality differs from uninjured LAPNs. For cluster 1 LAPNs, ventral (Mann-Whitney test, p=.022, sham 0.19 ± 0.09 vs. SCI 0.22 ± 0.11) neurite orientation differs post-SCI (G). For cluster 2 LAPNs dorsal (Mann-Whitney test, p<.0001, sham 0.11 ± 0.16 vs. SCI 0.16 ± 0.12), ventral (Mann-Whitney test, p=.0009, sham 0.16 ± 0.12 vs. SCI 0.20 ± 0.14), and medial (Mann-Whitney test, p<.0001, sham 0.31 ± 0.13 vs. SCI 0.22 ± 0.14) neurite orientations differ post-SCI (H). For cluster 3 LAPNs dorsal (Mann-Whitney test, p<.0001, sham 0.22 ± 0.15 vs. SCI 0.32 ± 0.19), ventral (Mann-Whitney test, p<.0001, sham 0.34 ± 0.19 vs. SCI 0.45 ± 0.22), medial (Mann-Whitney test, p=.0051, sham 0.14 ± 0.09 vs. SCI 0.11 ± 0.11), and lateral (Mann-Whitney test, p<.0001, sham 0.30 ± 0.18 vs. SCI 0.12 ± 0.11) neurite orientations differ post-SCI (I). Large center lines show group means and error bars are standard deviations. *p<.05, **p<.01, ***p<.001, **** p<.0001.

Neurite directionality of the largest and most complex LAPNs, cluster 1, was similar between sham and SCI, with only a modest increase in the percentage of ventral oriented neurites in the SCI group (Figure 5G). Cluster 2 LAPNs in both the sham and SCI groups oriented most neurites mediolaterally. However, after SCI cluster 2 LAPNs have a higher percentage of dorsally and ventrally oriented neurites and a lower percentage of medially oriented neurites (Figure 5H). Cluster 3 LAPNs in both groups have neurites predominantly oriented dorsoventrally. After SCI, there was a higher percentage of neurites oriented both dorsally and ventrally, and a lower percentage oriented medially and laterally (Figure 5I). As there were minimal differences in other morphometrics between sham and SCI LAPN clusters the differences in neurite directionality likely represent injury induced plasticity.

While unbiased cluster analyses showed that LAPNs form three distinct subsets based on morphology (Figure 2), and that neurite directionally between subsets differed between the groups, we directly compared sham and SCI LAPN morphology as singular sets of neurons (Supplementary figure 2). These analyses showed that LAPNs were largely similar between groups, with only soma volume (Supplementary figure 2C) and medial neurite directionally (Supplementary figure 1H) being decreased in the SCI group, which is similar to the previous analyses.

SCI alters LAPN axon collateral targets

Here, LAPNs have been defined as having somata at L2/3 and axon terminals at C5/6. However, this does not exclude the possibility that some LAPNs have axon collaterals targeting other spinal levels. As expected, in the intact cord LAPN axons/collaterals were seen at C6 (Figure 6A, top left panel) and were concentrated in lamina IX. Surprisingly, LAPN collaterals were also found at T9 (Figure 6A, top middle panel) and T12 (Figure 6A, top right panel) primarily in the intermediate gray matter. As LAPN axons ascend in the VLF, a large proportion likely remain intact following mild or moderate SCI, but the insult may induce changes in collateral densities and target areas. In the SCI group, collateral densities rostral to the injury decrease at C6 (Figure 6B). In the thoracic spinal cord collateral densities in the sham group were similar at T9 and T12 and a similar density was found in the SCI group at T9 (Figure 6C). However, there was a significant increase in collateral density caudal to the injury at T12 in the SCI group compared to all other spinal levels and groups (Figure 6C).

Figure 6.

5 weeks post-SCI LAPN axon collaterals densities differ by target. A. Heatmaps depicting LAPN axon densities at T9, T12, and C6. B. LAPN axon collaterals in the thoracic spinal cord differ post-SCI (Kruskal-Wallis test, p<.0001) with an increase in collaterals caudal to the injury at T12 (Dunn’s test, p<.0001, sham T12 3505 ± 2941 vs. SCI T12 6317 ± 3494; p<.0001 sham T9 4149 ± 4169 vs. SCI T12 6317 ± 3494; p=.0002 SCI T9 4149 ± 4169 vs. SCI T12 6317 ± 3494).C. LAPN axon collaterals in cervical spinal cord decrease at 5 weeks post-SCI (unpaired t-test, p=.03, sham 10827 ± 8171 vs. SCI 7259 ± 7087). Large center lines show group means and error bars are standard deviations. *p<.05, ***p<.001, **** p<.0001.

By definition, propriospinal neurons are wholly contained within the spinal cord. However, as there were fewer LAPN axons/collaterals at C6 in the SCI group, we anticipated that LAPN axons may target other motor centers post-SCI. Surprisingly, in both the sham and SCI groups, LAPN axons/collaterals were seen in the reticular formation (Figure 7A). Collaterals were quantified in the caudal gigantocellular reticular nucleus (Gi) and lateral paragigantocellular nucleus (LPGi), and no differences were seen between groups (Figure 7B, C).

Figure 7.

SCI does not alter LAPN axon collaterals in the reticular formation. A. Brainstem cross section stained with fluorescent Nissl (magenta) and viral labeled LAPN axon collaterals (green). White overlay is an anatomical map showing the gigantocellular nucleus (Gi) and lateral paragigantocellular nucleus (LPGi). B. Percent positive area for LAPN axon collaterals in the Gi are unaltered post-SCI. C. Percent positive area for LAPN axon collaterals in the LPGi are unaltered post-SCI. Large center lines show group means and error bars are standard deviations.

Discussion

LAPN morphology is heterogenous

Previous studies traced LAPNs in cats and rats, characterized the number or proportion of neurons that project ipsi- and contralaterally and qualitatively assessed morphology (Dutton, Carstens, Antognini, & Carstens, 2006; English et al., 1985; Molenaar, 1978; Reed et al., 2006). To briefly summarize, LAPNs were reported to project roughly 50% ipsilaterally (homolaterally) and 50% commissurally (heterolaterally), to have cell bodies in the deep dorsal horn and intermediate gray matter, and to have axons in the ventrolateral funiculus. They project to the intermediate and ventral gray matter of the cervical segments and are not known to have extensive collateralization at other spinal segments. In the current study we took advantage of robust viral-based tracing and tissue clearing techniques to complete a detailed quantification of LAPN soma and neurite morphologies, and axon/collateral distributions. We employed an unbiased hierarchical cluster analysis that grouped LAPNs into three distinct clusters based on soma size and/or dendrite orientation. Cluster 1 LAPNs had large somata with extensive neurite lengths and complexity. As total length and complexity of neurite arbors directly relates to the number of inputs, cluster 1 LAPNs likely receive more inputs than cluster 2 and 3 neurons. Additionally, larger somata tend to have lower input resistances and require more input to reach threshold potential (higher rheobase) and are associated with larger diameter, faster conducting axons (Carrascal, Nieto-Gonzalez, Torres, & Nunez-Abades, 2011; Harper & Lawson, 1985; Purves & Hume, 1981). These properties suggest that cluster 1 LAPNs are more likely to be recruited when faster conduction velocities are required for interlimb coordination during high-speed locomotion. Cluster 2 and 3 LAPNs had smaller somata and less extensive neurites compared to cluster 1 but differed in neurite directionality. Cluster 2 LAPN neurites were oriented mediolaterally suggesting that the primary inputs of these cells are axons from the lateral and ventrolateral white matter (Holmes & Berkowitz, 2014; Kathe et al., 2022), which includes the descending reticulospinal tract, vestibulospinal tract, and other propriospinal neurons - all of which are involved in locomotor output (Kjell & Olson, 2016). In contrast, cluster 3 LAPNs had neurites that oriented more dorsoventrally. Sensory afferents from the dorsal root ganglia enter the dorsal horn of the spinal cord and penetrate into the spinal gray matter in a dorsoventral direction. Thus, we speculate that cluster 3 LAPNs receive substantial input from sensory afferents and are involved in the integration and processing of incoming sensory signals (Bras, Cavallari, Jankowska, & Kubin, 1989; Rastad, Gad, Jankowska, McCrea, & Westman, 1990; Ritz & Greenspan, 1985).

The LAPNs studied here have neurites that primarily orient coronally, and thus have a modest rostrocaudal spread (z-projection depth within our 1.5 mm pieces of spinal cord). This morphologic characteristic seems to be conserved among other propriospinal populations as Deng et al. found the same characteristic in long descending thoracic PNs (Deng et al., 2016). In addition, in multiple species neurons located in the intermediate gray matter, where LAPN somata primarily reside, also have coronal neurite orientations (Berkowitz, Yosten, & Ballard, 2006; Gelfan, Kao, & Ruchkin, 1970; Sterling & Kuypers, 1967). This characteristic seems conserved and may represent a morphology that is driven by the circuitry architecture needed for the execution of movement without input from the forebrain (Grillner, 2021).

Minimal functional deficits following mild T10 contusive SCI

Clinically, most SCIs are contusions and even if classified as clinically complete there is some level of tissue sparing (NSCIS, 2016; Talbott et al., 2015). In the current study, we used a mild contusive SCI with high levels of spared white matter at the epicenter, including the VLF where LAPN axons ascend. The injury caused a transient drop in gross locomotor function and resulted in no detectable differences in sensory function. However, as BBB scoring is a crude measure of locomotor function, overground gait was assessed as a more precise analysis of locomotion. Mild contusive T10 SCI disrupted interlimb coordination of the homo- and heterolateral forelimb-hindlimb pairs but had no effect on the hindlimb or forelimb pairs or spatiotemporal gait indices. Previous studies have shown that morphologic changes in spared spinal neurons can correlate with function (Chew & Sengelaub, 2021; Deng et al., 2016). As we showed previously (Pocratsky et al, 2020; Shepard et al., 2021) LAPNs play a vital role in interlimb coordination pre-SCI and appear to impair the recovery of interlimb coordination post-SCI. Thus, it seems logical to speculate that the neuroanatomical changes seen here are in some way related to the functional deficits that are mild and restricted to interlimb coordination (Pocratsky et al., 2020; Shepard et al., 2021).

LAPNs differ morphologically following SCI

LAPN somata reside several spinal levels caudal to the injury and their axons ascend in the outer rim of the spared white matter, leaving many/most of them anatomically intact following contusive SCI. However, since injury-induced neuroplasticity is seen throughout the neuraxis and involves many neurons that are not themselves damaged, we hypothesized that LAPN morphology would differ post-SCI (Filli et al., 2014; Min et al., 2015). At 5 weeks post-SCI, hierarchal cluster analysis grouped LAPNs into three distinct clusters that were grossly similar to sham clusters (Figures 2A, 5A, Supplemental Figure 1), but with marked differences in neurite orientation. Following SCI, the neurites from neurons in clusters 2 and 3 were oriented more dorsoventrally. Post-SCI shifts in neurite orientation have also been reported for long descending thoracic PNs, but for these PNs the neurites assumed a more mediolateral orientation, opposite to the dorsoventral shift found here (Deng et al., 2016). Based on the principle that functional synaptic inputs influence neurite orientation, the shifts in both of these PN populations are likely a compensatory response to seek out new inputs and/or to facilitate the formation of new circuits post-SCI. Thus, the difference in orientation shift may be a consequence of soma location relative to the injury epicenter (Gray, Smith, & Rubel, 1982; Greenough & Volkmar, 1973; Harris & Woolsey, 1981; Katz & Shatz, 1996). The mediolateral shift in thoracic PNs, that are rostral to the injury is advantageous as descending PNs are involved in the formation of detour circuits in an effort to relay commands to targets caudal to the injury (Bareyre et al., 2004; Filli et al., 2014). However, descending input to LAPNs are disrupted post-SCI and the dorsoventral shift orients the LAPN neurites towards sensory afferents that are known to sprout caudal to the injury (Beauparlant et al., 2013; Krenz & Weaver, 1998). Interestingly, the changes in LAPN morphology closely resemble those seen in pyramidal neurons of the barrel cortex following deafferentation. In both cases, neurite orientation is altered but there are no changes in convex hull or total neurite length, suggesting that orientation differences are not just a consequence of neurite degeneration/loss but also of neurite growth in new directions. Such plasticity is referred to as homeostatic structural plasticity, in which synapses and resultant circuitry self-organize to assure network stability (Yin & Yuan, 2015). Furthermore, aspects of development are recapitulated post-SCI (Talifu et al., 2022), and LAPN neurites may go through the growth and stabilization phases of dendrite development. During the growth phase, the neurites increase in number, length and complexity and form NMDA receptor mediated synapses. Subsequently, active synapses show an increase in AMPA receptors. These neurites and synapses are maintained while neurites that fail to increase the ratio of AMPA/NMDA receptors are retracted (Dailey & Smith, 1996; Rajan & Cline, 1998; Wu, Zou, Rajan, & Cline, 1999). A shift to the stabilization phase then occurs when CaMKII expression increases (which occurs 3-14 days post-SCI) resulting in decreased new branch formation retraction (Dai et al., 2005). Given this, it is plausible that LAPNs are seeking out new functionally active inputs post-SCI in an effort to achieve homeostatic plasticity.

While there were significant differences in neurite orientation for LAPN clusters 2 and 3, the large complex LAPNs in cluster 1 showed little change after injury as only a modest increase in ventrally oriented neurites was seen in the SCI group. While this may seem counterintuitive given that larger neurons are often more susceptible to injury and disease (Frey et al., 2000; Hegedus, Putman, & Gordon, 2007; Sakurai, Hayashi, Abe, Sadahiro, & Tabayashi, 1998), we suggest that this reflects instead that the sources of input to cluster 1 LAPNs either did not change after injury, or that a neurite orientation shift would not be advantageous to the formation of new synapses. In our recent publications exploring synaptic silencing of LAPNs and the descending equivalent, the LDPNs (Pocratsky et al., 2020; Shepard et al., 2022; Shepard et al., 2021), we proposed that these two populations of interenlargement propriospinal neurons transmit temporal information, important for interlimb coordination, between the forelimb and hindlimb locomotor circuits. In this scenario the primary synaptic input onto some LAPNs would be local and likely dominated by output from the hindlimb central pattern generators (Shepard et al., 2022). Thus, these inputs would not be directly influenced by a thoracic contusive injury, which may be the case for cluster 1 LAPNs characterized here. Additionally, the seeming resistance to injury-induced plasticity of cluster 1 LAPNs may reflect differences in transcriptional profiles which alter a cells potential to respond to injury (He, Hirata, Wang, & Kawabuchi, 2003; Hu et al., 2016; Murray, Beauvais, Gibeault, Courtney, & Kothary, 2015). It would be interesting for future studies to compare the transcriptional profiles of cluster 1 LAPNs with cluster 2 and 3 LAPNs following SCI, as this may provide therapeutic targets.

LAPN collateral targets differ post-SCI

In both groups, LAPN collateral densities were greatest in lamina IX, but were also found throughout the intermediate gray matter of the C5 and C6 segments, including lamina X. The most straight-forward interpretation of these data is that LAPNs target motoneurons both pre- and post-SCI, but also that a smaller proportion target gray matter volumes (primarily laminae VII and X) where sensorimotor processing and motor pattern generation are key functions and where the majority of the corollary LDPN somata reside (Flynn et al., 2017; Reed et al., 2006).

Despite LAPN axons ascending in the intact VLF, LAPN collateral densities were lower at C5/6 in the SCI group. This may reflect a target disconnection from motoneurons at C5/6 as motoneurons rostral to the injury have been shown to die or atrophy post-SCI (Grumbles & Thomas, 2017; Huber et al., 2018). However, it could also reflect post-injury plasticity where synapse numbers decrease despite the numbers of source and target neurons remaining unchanged, as the source neurons (LAPNs) sprout into regions of thoracic spinal cord nearby the injury (see below). Here we have defined LAPNs as having somata at L2/3 and axon terminals at C5/6. However, we unexpectedly found collaterals in the gray matter of the thoracic spinal cord. Interestingly, LAPN collateral densities were highest in lamina VII at T9 and T12 in both groups (Figure 7A). While this is somewhat unexpected, it should be noted that our uninjured group were true sham animals, in that they received a laminectomy, but not a contusion injury. It is possible, though unlikely, that some of the collateral sprouting into the thoracic spinal cord was triggered by the local surgical procedure.

Throughout the length of the spinal cord the intermediate gray matter, Laminae VI, VII and X contain GABAergic premotor interneurons and spinocerebellar neurons, and is known to integrate sensory and motor inputs (Barber, Vaughn, & Roberts, 1982; Delwaide, Figiel, & Richelle, 1977; Krutki, Grottel, & Mrowczynski, 1998; Millan, 1999; Miller & Strominger, 1973; Skinner, Adams, & Remmel, 1980). This suggests that LAPNs likely contact ensembles of thoracic premotor neurons to coordinate and execute whole body movements such as locomotion. This coincides with electrophysiologic data from Juvin et al which show that activation and coupling of thoracic circuity ensures effective interenlargement coordination (Juvin, Simmers, & Morin, 2005). Furthermore, these findings provide anatomical substrates that at least partially explain the findings of Shepard et al, which showed that silencing LAPNs post-SCI only improved interlimb coordination of the hindlimbs but no longer had an impact on contralateral or forelimb interlimb coordination following injury (Shepard et al., 2021). The increase in collateral density caudal to the injury may represent aberrant plasticity that when silenced allows the lumbar circuitry to function appropriately via removal of excess ‘noise’ from the hindlimb locomotor system.

We further investigated whether LAPNs projected to supraspinal centers involved in locomotion. In the sham group, small numbers of LAPN collaterals were present in the gigantocellular (Gi) and lateral paragigantocellular (LPGi) nuclei of the reticular formation. Importantly, this means that this at least some of this population of “propriospinal” neurons are not wholly contained within the spinal cord, and therefore by definition are not truly propriospinal neurons. As there were differences in collateral densities at C6, we anticipated differences in the reticular formation as well. However, we found no post-injury differences in either the Gi or LPGi. As there were no differences in collaterals at T9, Gi, or LPGi, but a decrease at C6 and an increase at T12, it appears that LAPN collaterals are retracted from some target areas but retain or increase connectivity to other targets. Similar to the changes seen in neurite orientations, this is likely another example of homeostatic plasticity following injury.

Functional and clinical implications of LAPN plasticity

At 5 weeks post-SCI, conditional silencing of LAPNs improved gross locomotor function and normalized interlimb coordination (Shepard et al., 2021). These findings were counterinitiative, as silencing an anatomically intact pathway was expected to exaggerate locomotor deficits. However, our current data showing shifts in LAPN neurite orientation and differences in collateral densities post-SCI might help to explain these findings. Following an SCI, the CNS undergoes an intense period of injury-induced plasticity which includes changes in synaptic boutons, receptor densities, and circuit reorganization (Fauth & Tetzlaff, 2016). Formation and maintenance of synaptic connectivity is dependent on functional inputs, and LAPNs likely seek/receive new functional inputs post-SCI to maintain homeostasis. However, our findings and those of Shepard et al. (2021) suggest that post-SCI, at least some of the LAPN homeostatic plasticity is maladaptive. This is likely a consequence of: 1) disruptions in descending motor pathways that previously targeted LAPNs and 2) sprouting of sensory afferents into the dorsal horn/intermediate gray matter caudal to the injury. As there is little meaningful sensory and motor input to guide the formation of circuitry caudal to the injury during the intense period of acute post-injury plasticity, the resultant circuity is hyperexcitable and overactive which impairs the ability of the locomotor circuity to produce meaningful output (Kathe et al., 2022; Leem et al., 2010; Talifu et al., 2022). In agreement with this, Kathe et al. (2022) showed that epidural stimulation, which improves locomotor function post-SCI, actually decreased overall neural activity in the lumbar enlargement, but also resulted in an increase in activity of a PN subpopulation which are presumed to be responsible for the improved functional outcomes.

Conclusion

In this study we took advantage of robust and specific viral-based tracing and tissue clearing techniques to quantitively characterize ascending propriospinal neurons (LAPNs) with cell bodies at L2/3 and terminals/axons at C5/6. We found that these LAPNs are morphologically heterogenous. This heterogeneity suggests LAPNs have diverse synaptic inputs and potentially differing roles in locomotion. The coronal orientation of LAPN neurites is conserved among propriospinal neuron populations and species and represents a morphology that is advantageous for segmental control of locomotor output. Additionally, LAPN neurite orientation and collateral targets are altered post-SCI indicating that despite being anatomically spared, LAPNs exhibit SCI-induced plasticity which appears to be maladaptive. These findings are somewhat surprising given the very mild injury employed, reiterating that even mild SCI can impact neurons throughout the neuraxis and showing that there is marked anatomic plasticity caudal to the injury epicenter in the lumbar cord. This plasticity is likely triggered by a loss of descending inputs and increased input from sensory afferents which sprout caudal to the injury. Future studies may be able to detect more differences in these subpopulations using laser capture microdissection of individual LAPNs to characterize their transcriptomes pre- and post-SCI potentially identifying molecular therapeutic targets to treat maladaptive injury induced plasticity.

Limitations

The cluster analyses used here are unbiased and clustered LAPNs into distinct clusters both in the sham and SCI groups. For comparisons between sham and SCI groups, we compared individual clusters with one another. Though the morphometrics of the clusters were similar to one another, there is no way to confirm that the LAPNs in each cluster are part of the same subgroups of LAPNs in the sham and SCI groups. However, Supplementary figure 1 shows that there were no differences between the total number of labeled cells or the number of cells in each cluster pre- and post-SCI. Nociceptive input(s) are known to play a role in maladaptive plasticity (Detloff et al., 2016; Henneman, 1957) and animals underwent 2 different nociceptive tests at 5 weeks post-SCI. While these tests were only performed once post-SCI it cannot be ruled out that testing may have some influence on post-SCI plasticity.

Highlights.

• Dual-viral labeling allows for large volume single neuron reconstructions

• Long ascending propriospinal neurons have three distinct morphologies

• Spinal cord injury results in shifts in neurite orientation

• Long ascending propriospinal neurons’ axon targets change post-spinal cord injury