Reawakening inflammation in the chronically injured spinal cord using lipopolysaccharide induces diverse microglial states

Abstract

Background: Rehabilitative training is an effective method to promote recovery following spinal cord injury (SCI), with lower training efficacy observed in the chronic stage. The increased training efficacy during the subacute period is associated with a shift towards a more adaptive or proreparative state induced by the SCI. A potential link is SCI-induced inflammation, which is elevated in the subacute period, and, as injection of lipopolysaccharide (LPS) alongside training improves recovery in chronic SCI, suggesting LPS could reopen a window of plasticity late after injury. Microglia may play a role in LPS-mediated plasticity as they react to LPS and are implicated in facilitating recovery following SCI. However, it is unknown how microglia change in response to LPS following SCI to promote neuroplasticity.

Main body: Here we used single-cell RNA sequencing to examine microglial responses in subacute and chronic SCI with and without an LPS injection. We show that subacute SCI is characterized by a disease-associated microglial (DAM) signature, while chronic SCI is highly heterogeneous, with both injury-induced and homeostatic states. DAM states exhibit predicted metabolic pathway activity and neuronal interactions that are associated with potential mediators of plasticity. With LPS injection, microglia shifted away from the homeostatic signature to a primed, translation-associated state and increased DAM in degenerated tracts caudal to the injury.

Conclusion: Microglial states following an inflammatory stimulus in chronic injury incompletely recapitulate the subacute injury environment, showing both overlapping and distinct microglial signatures across time and with LPS injection. Our results contribute to an understanding of how microglia and LPS-induced neuroinflammation contribute to plasticity following SCI.

Keywords: Chronic nervous system injury; Inflammation; Lipopolysaccharide; Microglia; Plasticity; Single-cell RNA sequencing; Spinal cord injury.

Fig. 1

Microglia exhibit transcriptomic differences in subacute and chronic SCI, and with an LPS injection. A Schematic describing the experimental workflow and timeline. Female rats underwent a cervical SCI and animals were perfused, their spinal cords extracted, cold dissociated into a single cell suspension, sorted using Cd11b and then sequenced. B Microglia from the different libraries projected onto a UMAP, including 3,415 microglia from naïve animals, 14,767 from rats with subacute SCI, 1,188 from rats with chronic SCI and a saline injection (ChSaline), and 8,970 from rats with chronic SCI and an LPS injection (ChLPS). C Microglia (28,340) were clustered using Seurat and SCCAF, projected onto a UMAP and manually annotated. D Top DEGs for the different clusters. Homeostatic microglia are defined by the upregulation of canonical microglial genes (P2ry12, Cx3cr1), primed microglia by the expression of ribosomal genes (Rps29 and Rpl17), myelin transcript-enriched microglia (MTEM) by the presence of transcripts from other cell types (Plp1 and Gfap), DAM by the expression of Lgals3, Apoe and distinguished based on the presence of Spp1 (Spp1-DAM) and complement genes (Complement-DAM), Ccl3/4-microglia express Ccl3 and Ccl4, IRM express Ifit2 and Ifit3, and proliferative microglia express Mki67 and Birc5. Clusters of TM express a mixed phenotype between two microglial states. E The raw cell counts that make up each of the microglial clusters across the libraries. F Cells were downsampled to 1,188 cells and expressed as a percentage making up each cluster

Fig. 2

Microglia shift from primed to injury-induced and homeostatic states. A Microglial trajectories identified by scVelo based on the presence of unspliced and spliced RNA. B A plot showing terminal states identified by CellRank kernel. The terminal states are observed on the outside of the plot and the initial, transitioning states are observed on the inside of the plot. Microglial metabolic pathways identified by the Compass pipeline from the (C) subacute, (D) ChSaline, and (E) ChLPS microglia compared to microglia from naïve, uninjured spinal cords. Reactions are defined by Recon2 pathways and the transparency of the dot represents the statistical significance. A Wilcoxon rank sum test was used to test between the groups and a Cohen’s d statistic is computed for each reaction. The metabolic activity in the different (F) libraries and (G) clusters as identified by the Metabolic Landscape pipeline. Scores are a ratio, in which 1.0 represents mean activity for the microglial activity, scores above this represent upregulation in those groups and scores below represent downregulated groups. H The expression of genes in the metabolic pathways identified in (F, G) projected onto a UMAP

Fig. 3

Interactions between microglia and neurons shift from subacute to chronic SCI. Significant predicted CellChat interactions between microglia (blue) and neurons (red) at the (A) library and (B) cluster level. More probable interactions are a darker hue. C The genes encoding the predicted microglial ligands projected onto a UMAP. D An alluvial plot showing key GO terms associated with each of the different microglial clusters

Fig. 4

Injection of LPS augments DAM population in degenerated tracts caudal to the injury site. A Representative sagittal images of dorsal white matter caudal to the injury site in naive, subacute, ChSaline and ChLPS animals with Iba1 (green) labelling microglia and Gal3 (magenta) demarcating DAM (n = 6 per group). B Quantification of the relative densities of Iba1 + cells, Gal3 + cells and colocalized cells. C The percentage of Iba1 + cells that also express Gal3 + . Statistics run were a two-way ANOVA with a Tukey post hoc test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Mean and SEM are represented on graphs