Microbial production of short-chain fatty acids attenuates long-term neurologic impairment after traumatic brain injury

Abstract

Background: Traumatic brain injury (TBI) triggers persistent gut microbiome dysbiosis characterized by depletion of short-chain fatty acid (SCFA)-producing bacteria. However, the link between SCFA depletion and long-term neurologic impairment (LTNI) after TBI remains unclear. Previously, we and others noted the involvement of metabolite-sensing receptors and SCFA ligands in mouse models of neurodegenerative diseases, including Alzheimer's. Here, we further investigated SCFA-mediated neuroprotection in LTNI at both microbiome and single-cell resolution using the controlled cortical impact (CCI) model of TBI with a high-yielding SCFA diet to examine their mechanistic role in pathogenesis.

Methods: C57BL6/J mice were randomized to CCI (6 m/s, 2 mm) or sham surgery. Following surgery, mice were randomized to a study diet based on a balanced modification of the AIN93-G diet containing either 15% high amylose maize starch (HAMS) control diet or acetylated and butyrylated HAMS (HAMSAB) for 6 months to model increased SCFA production by bacterial fermentation in the gut. Morris water maze test and nesting assessment were performed at 1, 3, and 6 months after injury. The longitudinal gut microbiome changes were investigated by 16 S rRNA amplicon and metagenomic sequencing of fecal pellets at baseline, 1 month, and 6 months post-injury. At 6 months, pericontusional tissue was collected for single-cell RNA-sequencing following the 10X Genomics protocol or histologic analysis.

Results: Compared to the HAMS control diet, HAMSAB diet remodeled the CCI murine gut microbiome at an early phase, increased various SCFA-producing taxa, and attenuated neurologic deficits up to 6 months after CCI. In mice fed HAMSAB diet, single-cell transcriptomics and pathway analysis identified the promotion of neurogenesis, including increased doublecortin-positive immature neurons. In myeloid cells, HAMSAB induced an anti-inflammatory phenotype, inhibiting pro-inflammatory signaling interaction such as midkine signaling, and promoted differentiation to disease-associated microglia (DAM). Simultaneously, SCFAs reduced neurodegenerative pathway activity in neurons and glial cells and reduced phosphorylated tau deposition in pericontusional cortex.

Conclusions: Diet-facilitated microbial production of acetate and butyrate attenuates behavioral deficits of LTNI after TBI and produces enduring benefits at the single-cell level on the neuro-inflammatory and neuro-progenitor responses. This therapeutic approach could have a broader potential to prevent neurodegenerative disease.

Keywords: Microbiome; Neuro-inflammation; Neurodegeneration; Single-cell sequencing; Traumatic brain injury.

Fig. 1

SCFA-enriched diet improves Morris water maze performance after traumatic brain injury.A, Stool SCFA concentration in Sham_HAMS, CCI_HAMS, and CCI_HAMSAB groups at 1 month, 3 months, and 6 months after controlled cortical impact (CCI) (n = 8–10 mice/group). B, Morris water maze testing of the three groups at three-time points post CCI (n = 8–10 mice/group). Vis, visible platform trials. *, p < 0.05 between Sham_HAMS vs. CCI_HAMS; #, p < 0.05 between Sham_HAMS and CCI_HAMSAB groups. C, Probe trial results of the three groups at 1 month, 3 months, and 6 months after CCI. D, Nest building test results of three groups at 6 months post CCI (n = 8–10 mice/group). Ns, no significance. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001

Fig. 2

SCFA-enriched diet increases SCFA-producing taxa in murine gut microbiome. A, α-diversity comparison between CCI_HAMS and CCI_HAMSAB group at three-time points based on 16 S rRNA data (n = 8–10 mice/group). x axis, Before, before CCI modeling. 1 M, 1 month post CCI; 6 M, 6 months post CCI. y axis, Shannon index. *, p < 0.05. B, Principal coordinate analysis between groups based on the Bray-Curtis distance estimated in metagenomics data (n = 4 mice/group). C, Twelve variable genera between CCI_HAMS and CCI_HAMSAB groups after 6 months. D. Differential species between CCI_HAMS and CCI_HAMSAB after 6 months

Fig. 3

SCFA-enriched diet alters cell frequency and gene expression in pericontusional tissue.A, UMAP with cell type annotation. B, Cell type proportion within each sample. C, Pairwise transcriptomic difference quantification within each cell type, each meta-sample used in the Bhattacharyya distance calculation represents a subsample of 500 cells from principal component analysis for group comparison, and a sample of 500 cells regardless of cell types (random) as background for significance estimation. The statistical significance was evaluated by 100 times iterations. The fold change of the Bhattacharyya distance between the pairwise comparison and the random shows the transcriptome difference. D, Comparison of G1 phase cell ratio in each cell type between groups. E, Homeostatic index heatmap for intercellular communication strength of each cell type. Red represents increased similarity of the CCI_HAMSAB group to Sham_HAMS relative to CCI_HAMS. Blue represents decreased similarity of the inter-cellular communication between CCI_HAMSAB and Sham_HAMS relative to CCI_HAMS. White means no strength difference among the 3 groups. F, The concept plot of homeostasis index of inter-cellular communication strength similarity/difference. Ns, no significance. *, p < 0.05; **, p < 0.01; ***, p < 0.001

Fig. 4

HAMSAB attenuates post-traumatic neuronal apoptosis pathways and promotes neurogenesis gene expression.A, Violin plot of neuron apoptotic pathway’ activity comparison. ***, p < 0.001. ns, no significance. B, UMAP of re-clustered neuron with subtype annotation. C, Differentiation potency UMAP. Darker red indicates greater differentiation potency. D, Immature neuron-related pathway’s AUC (activity) UMAP. Red means high activity. Green means low activity. E, Gene ontology biological processes (GOBP) analysis UMAP of commissural neuron axon guidance. Red indicates high activity and green indicates low activity. F, Heatmap of WGCNA module-enriched GO pathway activity. The left module annotation represents the WGCNA gene module, and the pathways within each module that were enriched based on GO database pathway analysis. The top annotation is the subtype proportion in each group and the subtype frequency’s statistical significance was evaluated by the Chi-square test. G, The subtype-specific regulon ranked by regulon specificity score (RSS), the red dots represent the specific regulons. H, Heatmap of regulon module activity, the modules are shown in Fig. 4I. I, Left, the neuron regulon module clustered based on connection specificity index, modules from top to bottom are modules 1 to 6. Middle/right, the CellChat communication signaling pathway/the GO pathway enriched by the corresponding module’s regulon-regulated downstream activated or repressed genes. The pathway enriched by activated genes is in red, and blue for the pathway enriched from repressed genes

Fig. 5

HAMSAB decreases neurodegenerative disease-associated pathways and phosphorylated tau deposition after TBI.A-B, Pairwise comparison of neurodegenerative disease-related pathway activity in neurons across groups, assessed using the Wilcoxon rank-sum test. C, Left, IHC staining of AT8, a mouse monoclonal antibody targeting hyper-phosphorylated tau protein, in the pericontusional cortex of murine Sham_HAMS, CCI_HAMS, and CCI_HAMSAB groups at 6 months post-injury. Right, Quantification of the percentage of IHC-positive area across groups. D, Comparison of HDAC pathway activity in neurons across the three groups. E, Pathway activity comparison in astrocyte (ASC) and oligodendrocyte (OLG). Ns, no significance. *, p < 0.05; **, p < 0.01; ***, p < 0.001. F, Dopaminergic neurons’ pseudo-time-related differential genes between CCI_HAMS and CCI_HAMSAB groups. The first column represents the genes’ expression along pseudo-time in the 2 groups respectively. The 2nd column is the expression difference at each time point. The last column represents the mean expression difference of the gene along the pseudo-time. G, The GO pathway enrichment dot plot from genes in Fig. 5F

Fig. 6

HAMSAB induces anti-inflammatory features of myeloid cells in pericontusional tissue 6 months after TBI.A, UMAP of re-clustered myeloid cells with subtype annotation. B, Violin plot of subtype’s canonical marker gene expression. C, Macrophage-specific pathway activity’s AUC UMAP. Red means high activity, and green means low activity. D, Heatmap of WGCNA module-enriched GO pathway activity. The left module annotation represents the WGCNA gene module, and the pathway in each module was enriched from the gene module referred to by the GO database. The top annotation is the subtype proportion in each group and the subtype frequency’s statistical significance was evaluated by the Chi-square test. E, Violin plot for microglia activation pathway activity pairwise comparison. ***, p < 0.001. F, The ratio of M2 phenotype and M1 phenotype from each sample. Statistical significance was evaluated by ANOVA. *, p < 0.1; ns, no significance. G, CellChat intercellular midkine signaling pathway strength circle plot. The chord color represents the signaling outgoing cell type and the width means the strength. H, The disease-associated microglia (DAM) score of each microglia subtype. Statistical significance was evaluated by the Wilcoxon rank-sum test. *, p < 0.05 in CCI_HAMS and CCI_HAMSAB comparison. #, p < 0.05 in Sham_HAMS and Young_Normal comparison. I, Differentiation lineage branch prediction by Slingshot for the 3 groups. J, RNA velocity UMAP for myeloid cells, the arrow points to differentiation direction. K, Lineage 2 microglia’s pseudo-time-related differentially expressed genes between CCI_HAMS and CCI_HAMSAB groups. The first column represents the genes’ expression along pseudo-time in the 2 groups respectively. The 2nd column is the expression difference at each time point. The last column represents the mean expression difference of the gene along the pseudo-time. The genes were clustered by K-means with auto-selected optimal cluster numbers. L, The GO pathway enrichment tree plot from clusters 3 and 5 genes in Fig. 6K