Single cell atlas of spinal cord injury in mice reveals a pro-regenerative signature in spinocerebellar neurons

Abstract

After spinal cord injury, tissue distal to the lesion contains undamaged cells that could support or augment recovery. Targeting these cells requires a clearer understanding of their injury responses and capacity for repair. Here, we use single nucleus RNA sequencing to profile how each cell type in the lumbar spinal cord changes after a thoracic injury in mice. We present an atlas of these dynamic responses across dozens of cell types in the acute, subacute, and chronically injured spinal cord. Using this resource, we find rare spinal neurons that express a signature of regeneration in response to injury, including a major population that represent spinocerebellar projection neurons. We characterize these cells anatomically and observed axonal sparing, outgrowth, and remodeling in the spinal cord and cerebellum. Together, this work provides a key resource for studying cellular responses to injury and uncovers the spontaneous plasticity of spinocerebellar neurons, uncovering a potential candidate for targeted therapy.

Fig. 1. Single nucleus RNA sequencing of the lumbar spinal cord after thoracic contusion.

a Schematics depicting the experimental design for snRNA-seq, showing the injured thoracic cord and lumbar cord (with dark red representing the lesion) as well as nuclei isolation from the intact lumbar cord followed by droplet-based barcoding for single nucleus RNA sequencing. b Top, an overview of experimental design for injury and tissue collection. The lumbar spinal cord of three animals from each time point: uninjured, 1 dpi (day post injury), 1 wpi (week post injury), 3 wpi, and 6 wpi. c Uniform manifold approximation and projection (UMAP) visualization of 67,903 nuclei from uninjured and injured lumbar spinal cords, revealing 9 classes and 39 subtypes. Colored by green (neurons), yellow, astrocytes, orange-red (microglia), purple (OPCs), blue (oligodendrocytes), light blue (Schwann), light pink (pericytes), pink (ependymal), magenta (leptomeninges), brick-red (endothelial). d Multiplex immunohistochemistry (IHC) of the lumbar spinal cord from uninjured, 1 wpi and 3 wpi. Tissue was stained for NeuN (green), GFAP (yellow), IBA1 (red), TMEM119 (dark orange), and OLIG2 (blue). Scale bars are 200 µm (main) and 50 µm (inset), respectively. e Quantification of the proportion of cell types from the snRNA-seq data and immunohistochemistry in tissue. Mean ± SEM; snRNAseq, N = 3; immunohistochemistry, N = 4. Source data are provided as a Source Data file.

Fig. 2. Cell type composition in the uninjured spinal cord and after injury.

a–d UMAPs depicting subclustering of major cell types: a neurons, b astrocytes, c microglia/hematopoietic cells, d oligodendrocyte lineage and Schwann cells. e A bar plot showing the 39 cell types in the atlas and their relative percent in each sample in the uninjured spinal cord. Individual replicates (N = 3) are shown as well as mean ± SEM. f The relative composition of the 39 cell types comparing injured samples (1 dpi, 1 wpi, 3 wpi, and 6 wpi) to uninjured, generated using scCODA showing the final parameter output from scCODA (confidence interval shown as 3–97% high-density index around the mean). Cell types with an inclusion probability > 0.85 were deemed significant. Significance depicted with a red asterisk. N = 3. For the log FC of individual replicates, see Supplementary Fig. 8. Source data are provided as a Source Data file.

Fig. 3. Plasticity-related expression in neurons after injury.

a Schematic depicting lumbar spinal cord neurons and their response to injury, whether that be an ascending neuron or an interneuron. b Immunohistochemistry of the lumbar spinal cord from uninjured, 1 wpi and 3 wpi. Tissue was stained for neurofilament (a cocktail of neurofilament-light, neurofilament-medium, neurofilament-heavy; purple) and MAP2 (green). Scale bars are 200 µm. c Quantification of neurofilament and dendritic changes. Pixels were quantified from thresholded images of neurofilament and MAP2. Error bars are mean ± SEM (N = 4). d Differential gene expression analysis comparing uninjured to 1 and 3 wpi neurons, the time points of maximal neuronal gene expression changes. Black dots indicate p value adjusted < 0.001, gray indicate p ≥ 0.001. e Pathway analysis for differentially expressed genes between uninjured and injured time points. X-axis indicates −log(p val adj) of GO and KEGG pathway clusters. P values (adjusted) were calculated using Benjamini–Hochberg false discovery rate (FDR). Yellow indicates relatively high normalized average expression and dark blue indicates relatively low normalized average expression. f Chord plot indicating shared genes between top 5 GO terms from genes upregulated 1 and 3 wpi. g Heatmaps showing average neuronal gene expression from top GO terms, including neurotransmitter receptors, synaptogenesis, and synaptic structure. Yellow indicates relatively high normalized average expression (1) and dark blue indicates relatively low normalized average expression (−1).

Fig. 4. Specific neurons express genes associated with regeneration.

a UMAP showing predicted neuronal families. b Targeted view of RAG-expressing cluster over injury time points. c Featureplots showing RAGs expressed in neurons. d A dotplot showing expression of RAGs within targeted RAG-expressing cluster (cluster 23 defined by independent clustering in Supplementary Fig. 5). Average expression (avg exp) is indicated by color (gray to blue) and percent expressed (pct exp) is indicated by the size of the circle. e, f RNAscope in situ hybridization showing expression of Sprr1a and Atf3 in the uninjured cord and 1 wpi. Scale bars are 200 and 50 µm, respectively. g Quantification of Sprr1a+ cells and Atf3+ cells in the uninjured and 1 wpi injury cord (p = 0.001 shown as ***p = 0.0037 shown as **, two-sided unpaired t-test, Error bars indicate ± SEM, N = 7, 10). h Diagram of spatial location of transcription types, including ChAT (light green), Vsx2 (dark green), and Shox2 (orange). i–l Visualization and quantification of VGluT2/Slc17a6+, Shox2+, Chx10/Vsx2+, and Sprr1a-expressing cells in the ventral spinal cord. Scale bars are 50 µm. Error bars indicate ± SEM (N = 5, 4, 4, 6 animals). m Diagram of spatial location of connectivity types, including ascending neurons labeled by dextran (aqua) and spinocerebellar (SCT, orange) neurons. n Visualization and quantification of lumbar spinal cord neurons labeled by dextran injected into a thoracic contusion lesion site. Aqua arrows indicate ATF3 and dextran overlap. Scale bars are 50 µm. Error bars indicate ± SEM (N = 5 animals). Visualization and quantification of the percent of RAG-expressing cells—Sox11 (o), Sprr1a (p), and ATF3 (q) that are spinocerebellar. Spinocerebellar neurons are shown in green and RAGs are shown in red. Orange arrows indicate RAG gene and spinocerebellar dual-labeled cells. Scale bars are 50 µm. Error bars indicate ± SEM (N = 4, 4, 5 animals). Source data are provided as a Source Data file.

Fig. 5. Spinocerebellar neurons express RAGs and display thoracic sprouting after injury.

a Schematic of AAV- injection and injury. b–d Spinocerebellar neurons and their cell bodies, axons, and mossy fibers in the cerebellum, thoracic and lumbar spinal cord. Virus expression is shown in green. Scale bars are 500 µm in the cerebellum and 200 µm for spinal cord sections (thoracic and lumbar). e Quantification of dendritic arborizations in SCT neurons (ns, p = 0.206, p = 0.211, Mann–Whitney test). f Quantification of thoracic axons rostral (R) and caudal (C) to the injury site (p = 0.0012, p = 0.035, Mann–Whitney test). g Quantification of gray matter collaterals rostral (R) and caudal (C) to the injury site, as measured by pixels after thresholding (p = 0.0485, two-way ANOVA). h Quantification of mossy fibers, normalized by the number of SCT neurons in the same animal (p = 0.336, p = 0.039, two-sided unpaired t-test). Mean ± SEM, N = 4, 5, 6 animals. *p < 0.05; ****p < 0.0001; n.s. not significant. Source data are provided as a Source Data file.