Short Chain Fatty Acid Supplementation After Traumatic Brain Injury Attenuates Neurologic Injury Via the Gut-Brain-Microglia Axis

Abstract

BACKGROUND:

Traumatic brain injury (TBI) is an underrecognized public health threat. There are limited therapeutic options for TBI, and supportive care remains the mainstay of treatment. Our previously published data demonstrate that post-TBI fecal microbiome transplantation (FMT) can reverse TBI-induced depletion of commensal bacteria, preserve white matter connectivity and neurocognition, and decrease cortical volume loss in mice after TBI.

HYPOTHESIS:

We hypothesized that post-TBI supplementation with Short Chain Fatty Acids (SCFAs), metabolites of commensal gut bacteria, would attenuate neurologic injury after TBI in mice.

METHODS:

14-week-old male C57BL/6 mice (n=52) underwent TBI via a controlled cortical impact vs. sham injury. Post-TBI, each group was treated with the SCFAs acetate, butyrate, and propionate vs. molar equivalent sodium chloride vehicle via free access to drinking water for four weeks post-TBI. The stool was collected three days pre-and sixty days post-TBI to assess the gut microbial community structure via 16s ribosomal RNA gene amplicon sequencing. Neurocognitive testing was performed with open-field and zero-maze testing. Ventricular volume and white matter connectivity were measured with 3D, contrast-enhanced MRI. Lastly, the transcriptional response of microglia was assessed with single-cell RNA sequencing (scRNAseq).

RESULTS:

SCFA supplementation decreased TBI-induced microbial loss, attenuated ventricular volume loss, preserved white matter connectivity, and altered the transcriptional profile of microglia after TBI. Post-TBI SCFA supplementation preserved the abundance of the butyrate-producing taxa Firmicutes, Clostridia, Ruminoccacaceae, and Peptoccacaceae (p=0.01). SCFA also reduced the TBI-induced increase in Clostridiales and Bacteroidales compared to the salt vehicle group (p=0.05). We also observed the preservation of non-TBI murine anxiety-like behavior in SCFA-treated TBI mice compared to vehicle-treated TBI mice in zero-maze (152.3 ± 101.8 cm vs. 147.5 ± 60.0 cm, p=0.006). These results were recapitulated with open field testing (11.7 ± 3%-time in the center in SCFA-treated TBI mice vs. 15.0 ± 6% %-time in the center of the field in vehicle-treated mice; p=0.002). Lastly, we observed upregulation of transcripts for the neuroprotective heat shock family of proteins and downregulation of neurodegeneration-associated transcripts, indicating an overall neuroprotective phenotype in microglia after SCFA supplementation post-TBI.

CONCLUSIONS:

We hypothesized that SCFA supplementation would attenuate neurologic injury after TBI in mice. SCFA supplementation attenuated neurocognitive deficits, reduced cortical volume loss, preserved white matter connectivity, and decreased neuroinflammation. These benefits may result from the direct replacement of SCFAs. However, there may also be secondary mechanisms related to commensal refeeding of butyrate-producing bacteria within the gut microbial community, a neuroprotective heat shock response, and a decrease in the expression of genes associated with neurodegeneration. The current study highlights the role of SCFAs in microbiome homeostasis and the potential of dietary intervention as a novel therapy in TBI.

Introduction

Traumatic brain injury (TBI) is a growing and under-recognized public health threat. The CDC estimates nearly 3 million people sustain a traumatic brain injury (TBI) each year in the United States, contributing to over 30% of all injury-related deaths ^1,2. TBI-related healthcare expenditures are nearly 80 billion dollars annually³. The impact of TBI is highlighted by its high mortality rate and by the significant long-term complications suffered by its survivors with the progressive development of motor, cognitive, and behavioral disorders^4-7. TBI triggers a robust pro-inflammatory response within the injured brain. The degree of this initial pro-inflammatory response has significant value in predicting long-term outcomes after TBI. Even after the acute inflammatory response has resolved, several studies have demonstrated residual long-lasting inflammation within the brain in animal models and patients ^4,8-10. One of the main drivers of this continued inflammation is the persistence of activated microglia, as microglia can remain activated for years after the initial insult. This immune response to TBI plays a fundamental role in developing and progressing subsequent neurodegenerative disease. It represents a complex interplay between the injured brain, the gut microbiome, and microglia—the resident innate immune cells of the brain ^10,11. TBI results in increased gut permeability, systemic inflammation, and changes within the gut microbiome. These changes, in turn, can alter the function within the brain^12,13. Indeed, preclinical evidence demonstrating the modulatory effects of gut microbiota on the brain has been shown in "germ-free" models where the disease is studied in animals devoid of microorganisms, models in which microbial manipulation is accomplished with antibiotics, and models of fecal microbiota transplantation (FMT)^13-15. These data suggest that the gut microbiota is mechanistically essential for regulating several CNS processes.

Recently published data from our laboratory shows that fecal microbiota transplantation restores gut microbial community structure and improves neurologic outcomes after TBI ^8,16,17. These data also show a marked preservation of cortical volume and white matter connectivity with FMT after TBI. These neuroanatomic findings correlated with significant differences in inflammatory gene expression within the microglia of FMT-treated TBI mice. Furthermore, the heat shock response, known to play a neuroprotective function in stroke and other neurologic disease processes, was enhanced within the microglia of FMT-treated mice compared to those receiving sham FMT after TBI. Though the mechanism(s) underlying the therapeutic benefit of FMT are multifactorial, the effects may, in part, be tied to the metabolic byproducts of the gut microbiota.

Multiple studies show TBI-induced depletion of several bacterial groups responsible for fermenting undigestible dietary fiber into short-chain fatty acids (SCFA) ^16-19. Other work has shown that altering the gut via probiotics and bacterial depletion leads to sex dependent effects on the brain gut axis after trauma ^20-22. SCFAs serve as signaling molecules via activation of G-protein-coupled receptors or inhibition of histone deacetylases. They regulate inflammation, glucose metabolism, blood-brain barrier maintenance, and microglia maturation and activation. For example, in stroke patients, decreased abundance of butyrate-producing bacteria and low fecal acetate levels are associated with poor functional outcomes. Likewise, in experimental models, repleting SCFAs after stroke has demonstrated improved neurocognitive outcomes ^23-27. These data, in line with our preliminary data, indicate that the gut microbiota and their metabolites play an essential role in the pathogenesis of neurologic injury. In the current study, we hypothesized that post-TBI dietary supplementation with SCFAs would improve neurologic outcomes after TBI. To test this hypothesis, we provided SCFAs versus a salt vehicle control within the drinking water of mice after TBI or sham injury. We then assessed anatomic, molecular, and pathologic outcome measures.

Methods

Animals

C57BL/6 male mice (Mus musculus) (28-30 grams) were procured from the Jackson Laboratory (Bar Harbor, Maine). Mice (n=52) were delivered at 12 weeks of age. We combined bedding and redistributed it into cages upon arrival to allow microbiome normalization. Mice were allowed two weeks of facility acclimation before the initiation of the experiment at 14 weeks of age. Housing and pathogen-free barrier maintenance were performed in Northwestern University's Center for Comparative Medicine facility. Ad libitum water and standard chow (Harlan, Indianapolis, IN) were provided for feeding. Animals were housed on a 12:12 light-dark cycle for the study duration. Mice were treated per the National Institutes of Health Guidelines for the Use of Laboratory Animals. At the experimental endpoint, euthanasia was performed following AVMA recommendations via carbon dioxide inhalation followed by exsanguination and decapitation. Brains were then collected for postmortem analyses. The Northwestern University Institutional Animal Care and Use Committee approved the experimental protocol.

Experimental Design

We employed a 2 x 2 study design to determine the interaction between post-injury SCFA supplementation (SCFA vs. vehicle) and traumatic brain injury (TBI vs. sham). 12-week-old male, C57BL/6 mice (n=52) were randomly assigned into four experimental groups— 1) TBI/SCFA, 2) Sham/SCFA, 3) TBI/Vehicle, and 4) Sham/Vehicle. All mice underwent bed mixing two weeks before assignment to allow for the normalization of gut microbial community structure between experimental groups (Sup Fig. 1, http://links.lww.com/SHK/C597). Time points for specific tests were selected using the behavioral literature as a guide for observing the specified chronic effects of TBI and SCFA ²⁸. At 14 weeks of age, TBI or sham injury was induced. As we have previously published, TBI was induced via an open-skull controlled cortical impact ²⁹. Post-injury, sham, and TBI mice received ad libitum water containing short-chain fatty acids (sodium propionate (25.9mM), sodium butyrate (40 mM), and sodium acetate (67.5mM)) or a molar-equivalent salt (sodium chloride (133.4mM)) as a control for a total of 4 weeks post-injury. Primary factors in this treatment regimen included animal welfare, clinical applicability, and previous research^23,25-27. At 59 days post-injury, mice underwent 3D, contrast-enhanced, magnetic resonance imaging. At 60 days post-injury, one cohort of mice was sacrificed, and brains were harvested for pathologic analysis. The second cohort was sacrificed, brains were harvested and microglia sorted via for Fluorescence-Activated Cell Sorting (FACS) for single-cell RNA sequencing (scRNAseq).

Traumatic brain injury

All groups were anesthetized via intraperitoneal injections of 125 mg/kg ketamine (Ketaset, Fort Dodge, IA) and 10 mg/kg xylazine without oxygen supplementation following standard protocol (Anased, Shenandoah, IA). All mice received a traumatic brain injury via a controlled cortical impact, as we have previously published ^8,29,30. In brief, a 1-cm scalp incision was made with a scalpel to reveal the sagittal and occipital sutures of the skull. The injury site was marked 2 mm rostral to the occipital suture and 2 mm left of the sagittal suture. A 5mm craniectomy was performed, leaving the dura mater undisturbed. The head was stabilized in a stereotaxic operating frame. A commercially available impacting device (Impact One, Leica Biosystems, Des Planes, IL) was used to administer a controlled cortical impact with a 3mm impacting rod at a velocity of 2.5 m/s and a depth of 2 mm²⁹. Sham mice received anesthesia and scalp incision only as craniectomy alone can result in a mild traumatic brain injury in mouse models. After a sham injury or TBI, scalp incisions were closed using VetBond (3M) (Santa Cruz Animal Health, Dallas, TX). All animals received a preinjury dose of buprenorphine SR (SR Veterinary Technologies, Windsor, CO) for post-procedure analgesia. Animal groups recovered in separate cages over a warming pad until fully mobile.

Short Chain Fatty Acid Administration

On the day of injury, short-chain fatty acids were dissolved into a liter of drinking water by volume. Short-chain fatty acids included sodium propionate (25.9 mM), sodium butyrate (40 mM), and sodium acetate (67.5 mM) ^25,31,32. A sodium chloride (133.4mM) molar equivalent was added to the water as vehicle control for sham treatment groups. Water and standard chow were provided ad libitum to the animals for four weeks post-injury. Water bottles were refilled every 3-4 days on average.

Behavioral Phenotyping

All mice(n=8/group) underwent behavioral phenotyping starting 45 days post-injury in the Northwestern University Behavioral Phenotyping Core during the light phase cycle, as we have previously published ^8,30. After a week, the studies were concluded. To strengthen the behavioral assessments, we included eight mice per group.

The Elevated Zero-Maze

Testing began with the elevated zero-maze (ZM), which began 45 days post-injury. As our lab has previously published, this test measures the time spent in open/closed quadrants of a circular maze to evaluate anxiety ^30,33. The animals were placed inside the center of the closed quadrant under white/yellow light. Their movements were tracked and recorded using Limelight 4 software (Actimetrics, Wilmette, IL). The maze was cleaned between mice using 70% ethanol.

Open Field

The open field (OF) test measured anxiety, exploratory behavior, and locomotive activity^30,33,34. As previously published, mice were placed into the center of the 54.5 x 54.5 cm square box with 30cm tall walls. During the test, the animals were exposed to white/yellow light. Each trial was 5 minutes, and movements were recorded and tacked using Limelight4 software (Actimetrics, Wilmette, IL). After each trial, the box was cleaned with 70% ethanol to minimize olfactory cues.

Fecal sample collection and characterization of microbial community structure

DNA extraction

Fecal samples were collected (59 days post-injury) and frozen at −80°C until DNA extraction. Genomic DNA was extracted from feces using bead-beating and automated purification using a Chemagic DNA Stool 200 Kit H96 (Revvity, Hamburg, Germany) implemented on a Chemagic 360 device. Samples (1-2 pellets per tube) were initially placed in ZR BashingBead Lysis Tubes (0.1 & 0.5 mm) (S6012-50; Zymo Research) with 1 ml lysis buffer and 20 μl proteinase K (both from the Chemagic DNA stool kit). Samples were subject to thermal mixing at 70°C for 10', followed by bead-beating using a TissueLyzerII device (Qiagen) set at 30 Hz for 30". A two-minute rest was provided in between two sequential bead-beating episodes. Subsequently, a five-minute incubation at 95°C was performed before final purification on the Chemagic360 instrument according to the manufacturer's suggestions.

Library preparation and sequencing.

Genomic DNA was PCR amplified with primers sIDTP5_515F and sIDTP7_806R (modified from the primer set employed by the Earth Microbiome Project (EMP; CTACACGACGCTCTTCCGATCTGTGTGYCAGCMGCCGCGGTAA and CAGACGTGTGCTCTTCCGATCTCCGGACTACNVGGGTWTCTAAT, respectively – underlined regions represent linker sequences) targeting the V4 regions of microbial small subunit ribosomal RNA genes. Amplicons were generated using a two-stage PCR amplification protocol with added linkers compatible with xGen^™ Amplicon UDI Primers (Integrated DNA Technologies, IDT)³⁵. First-stage PCR amplifications were performed in 10 microliter reactions in 96-well plates using RepliQa HiFi ToughMix (Quantabio). PCR conditions were 98°C for 2 minutes, followed by 28 cycles of 98°C for 10", 52°C" for 1" and 6 "°C for 1". Subsequently, a second PCR amplification was performed in 10 microliter reactions in 96-well plates using RepliQa HiFi ToughMix. Each well received a separate primer pair with unique dual barcodes from IDT (Catalog numbers: 10009846, 10009851, 10009852, and 10009853). One microliter of PCR product from the first stage amplification was used as a template for the 2nd stage, without cleanup. Cycling conditions were 98°C for 2 minutes, followed by eight cycles of 98°C for 10", 60°C" for 1" and 6 "°C for 1". Libraries were then pooled and sequenced with a 10% phiX spike-in on an Illumina Miniseq sequencer employing a mid-output flow cell (2x154 paired-end reads). DNA extraction, library preparation, pooling, and sequencing were performed at the Genomics and Microbiome Core Facility (GMCF) at Rush University.

Bioinformatics analyses.

Microbiome bioinformatics was performed using the software package QIIME2 2021.11³⁶. Raw sequence data were checked for quality using the software package FastQC and merged using PEAR³⁷. Merged sequences were quality filtered using the q2-demux plugin and denoising with DADA2 within QIIME^37,38. Primer sequences were removed using the Cutadapt algorithm. Taxonomy was assigned to ASVs using the q2-feature-classifier³⁹. Classify Sklearn-based naïve Bayes taxonomy classifier against the SILVA 138 99% reference sequences database. The contaminant removal software (Decontam) detected four contaminant ASVs based on the prevalence of the ASVs in the reagent negative blank controls using default parameters; these were removed from all samples before downstream analysis. Contaminants included three Mycoplasma ASVs, including M. wenyonii and Spingomonas. Alpha-diversity metrics (observed features, Shannon Index, Simpson's, and Pielou's) and Beta diversity metrics were calculated with QIIME after samples were rarefied (subsampled without replacement) to the same depth (17,000 sequences/sample) for all samples ⁴⁰. Permutational Multivariate Analysis of Variance (PERMANOVA) was conducted within the QIIME2 environment with 9,999 permutations at the taxonomic level of feature (i.e., ASVs) using Aitchison distance, a linear measure of sample dissimilarity for compositional data. Metagenomic pathways were inferred from 16S rRNA marker genes using the PICRUSt2 plugin within the QIIME2 environment ⁴¹. Generated pathways were annotated against the MetaCyc Metabolic Pathway Database ⁴². Bioinformatics analyses were performed at the Research Bioinformatics Core (RBC) at Rush University.

Tissue Harvesting for FACSorting

Thirty days after TBI or a sham injury, two brains per condition were digested in Liberase (Roche, Basel, Switzerland) and DNAase I (Roche, Basel, Switzerland). The samples were morcellated and placed into C-tubes (Miltenyi Biotec, Bergisch Gladbach, North Rhine-Westphalia, Germany). For FACS analysis, two randomly selected brains per condition were pooled to obtain sufficient cell yield and reduce individual biological variability. Sorting specifically targeted CD45⁺CD11b low microglia, gating out debris and other immune cells as previously reported ⁸. Samples were dissociated in C-tubes using a MACS dissociator (Miltenyi Biotec, Bergisch Gladbach, North Rhine-Westphalia, Germany). The samples were incubated for 30 min at 37 °C. The tissue mixture was strained through a 40 μm nylon mesh strainer and diluted with autoMacs Running Buffer (Miltenyi Biotec, Bergisch Gladbach, North Rhine-Westphalia, Germany). Residential and infiltrating cells were isolated using a 30/70 Percoll gradient (Percoll Plus; GE Healthcare, Chicago, IL, USA). The interface containing the immune cells was retrieved and then washed. Cells were then stained with Live/Dead Aqua (Invitrogen, Waltham, MA, USA) viability dye, Fc-Block (Biosciences, San Jose, CA, USA), and fluorochrome-conjugated antibodies for CD45.1(A-20/BD Biosciences, San Jose, CA, USA) and CD11b (M1/70/BD Biosciences, San Jose, CA, USA). The cells were then washed and sorted on a BD FACSAria cell sorter (BD Biosciences, San Jose, CA, USA). As we previously reported, gates were established using "fluorescence minus one" controls ⁴³.

MRI Acquisition and Image Processing Methods

At 59 days post-TBI or sham injury, mice underwent magnetic resonance imaging at the Northwestern University Center for Translational Imaging ⁴⁴. MRI Brain images were acquired and processed, and the region of interest was selected as previously published⁴⁵. A 3D multi-gradient echo sequence (mGRE) with the following parameters was used to obtain high isotropic resolution of each mouse brain (150 micrometers in each direction) as well as 3D R2* maps with the following MR timing and acquisition parameters: TR=80msec, and 12 echoes with TE=2.5, 6.3, 10.1, 13.9, 17.7, 21.5, 25.3, 29, 32, 36.7, 40.4, and 44.2 milliseconds) and FA=20. A 2D T1 map was generated using a gradient echo sequencing (GRE) sequence with multiple flip angles (5, 10, 15, 30, 40, 50, 60, 70, and 90). Finally, a 2D echo planar sequence with three orthogonal directions and multiple -values was used to obtain diffusion-weighted images and to generate modified fractional anisotropy (FA) patterns using a previously described method ⁴⁵ with an added contrast agent⁴⁶. A 3D scan was repeated using the same 3D sequence as described to enhance possible changes in ventricular volume associated with blood-brain-barrier disruption. Analysis of morphological data and extraction of quantitative parameters was accomplished using several different software packages: ImageJ (Xinapse) for generating quantitative maps and ITK-SNAP for threshold-based segmentation and quantifying ventricular volume.

Single-cell RNA-sequencing

Library preparation and sequencing

The concentration and viability (>90%) of FACSorted CD45⁺CD11b⁺ cells were confirmed using a K2 Cellometer (Nexcelom Bioscience LLC, Lawrence, MA) with AO/PI reagent, and ~5,000–10,000 cells were loaded on 10x Genomics Chip A using Chromium Single Cell V3 Reagent Kit and Controller (10x Genomics, Pleasanton, CA, USA). Single-cell 3' RNA-Seq libraries were prepared according to the manufacturer protocol (10x Genomics, Pleasanton, CA, USA). Libraries were assessed for quality (TapeStation 4200, Agilent Technologies, Santa Clara, CA, USA) and then sequenced on a HiSeq 4000 instrument (Illumina, San Diego, CA, USA), generating >25,000 read pairs/cell.

Preprocessing of scRNA-seq data

Raw data were processed using Cell Ranger, version 3.0, the pipeline from 10x Genomics (Pleasanton, CA, USA). The reads were mapped to the mm10 mouse reference genome (Ensembl Build 98). Individual sample expression matrices were loaded and read into R using the functions Read10x and CreateSeuratObject in the Seurat package version 5.0.1. Cells expressing less than 600 genes, less than 1400 unique molecular identifiers (UMIs), and more than 10% of mitochondrial genes were excluded. For quality control of unwanted variation due to cell cycle phases, all samples were scored using CellCycleScoring in the Seurat package. However, regression was not performed due to minimal cell cycle variations. Correlation analysis was conducted to test within-group variation as opposed to inter-group variation. Following SCTransform normalization, samples were integrated using the PrepSCTIntegration and IntegrateData functions to correct the batch effect between samples by 30 principal components. Twenty-seven clusters were identified using unsupervised clustering and a dimension resolution of 1. Further annotation of these clusters was performed using canonical markers from curated literature^8,26,47,48. The R packages ggplot2 (version 3.4.2) and Seurat (version 5.0.1) were used for all plots.

Analyses of scRNA sequencing data

Microglia were segregated using the Seurat subset function and re-integrated for unsupervised clustering using the RunPCA function with a principal component 5 and dimension resolution of 0.8. The AddModuleScore function was used on SCT objects to calculate the average expression levels of each group. For differential expression analysis of microglia, the FindMarkers function was used with the MAST package to identify DE genes between groups: TBI vs. Sham, SCFA vs. Sham, and SCFA_TBI vs TBI. Ribosomal, mitochondrial, and human leukocyte genes were removed to reduce noise. A custom volcano plot function was used to plot statistically significant DE genes with Bonferroni correction to represent DE analysis results visually. Downstream gene ontology (GO) enrichment analysis was performed using GOrilla with all genes within the cell type in the dataset as a background ⁴⁹. The false discovery rate (FDR) was computed to adjust for multiple testing scenarios. The results of the GO analysis were then visualized using the ggplot2 packages, highlighting FDR q-values and enrichment scores.

Statistical Analyses

Statistical significance among different groups was assessed using a two-way ANOVA. Multiple comparisons test was applied to refine the statistical analysis. Three-way ANOVA, accompanied by Šidák's comparisons test, was employed to evaluate statistical significance accounting for cell type, treatment, and injury. Data are presented as mean ± SEM. Group means were compared by analysis of variance (ANOVA) with Tukey post hoc analysis. For all analyses, the threshold for statistical significance was set at P < 0.05. ((p ≤ 0.05(*), 0.001(**), & 0.0001(***)). All analyses used GraphPad Prism (GraphPad, La Jola, Ca, version 6).

Results

SCFA Effect on Gut Microbial Community Structure

We and others have previously published alterations observed in the gut microbial community structure in response to TBI, specifically in bacterial taxa that produce SCFAs ^50,51. In our current model, the short-chain fatty acids acetate, propionate, and butyrate were supplemented within the drinking water of TBI and sham-injured mice for four weeks following TBI in a post-injury supplementation paradigm. The gut microbial community structure was then determined via 16s rRNA sequencing. For all experimental groups, the stool was collected three days before the induction of TBI and 59 days post-injury (DPI). While no statistically significant differences in alpha or beta diversity were observed among samples collected at day −3 post-injury (DPI), apparent differences emerged by day 59 following TBI (Fig. 1A-C). At this later time point, alpha diversity, measured using the Shannon index, showed a non-significant decrease in the TBI group (0.944) relative to the sham group (0.956), with ANOSIM yielding p = 0.89 and PERMANOVA p = 0.94 (Fig. 1A). In contrast, beta diversity assessed via a Bray-Curtis PCoA revealed statistically significant shifts in microbial community structure at day 59, as indicated by ANOSIM (p = 0.005) and PERMANOVA (p = 0.012) analyses (Fig. 1B). Subsequent differential sequencing analyses identified 11 key bacterial genera that contributed to the observed differences in community composition between TBI and sham groups (Fig. 1C). These groups have robust literature supporting their link to health and disease—Clostridia and Bacteroides have been linked to inflammation, Oscillobacter and Lachnospiracae with cognition and neurodevelopment, and Blautia, Lachnospiraceae, and Bacteroides with short-chain fatty acid production ^52,53. Most strikingly, the SCFA-TBI group demonstrated a bacterial composition much more akin to the sham animals than compared to the TBI-alone group.

Figure 1. Post-injury SCFA treatment mitigates TBI-induced dysbiosis within the gut microbial community structure.

Fecal pellets were collected from each experimental group (Sham, TBI, SCFA, SCFA-TBI) 3 days before and 59 days post-injury (DPI). Bacterial taxonomy was assessed using computational analysis of present 16s rRNA sequences. No statistically significant differences between samples at day −3 DPI in alpha or beta diversity were found. A. At day 59 post-TBI, the Shannon index revealed no significant differences in alpha diversity between experimental groups (Anosim, permanova group significance p=0.89, p=0.94). B. Bray principal components (PCoA) plot depicting the significant differences in beta-diversity distance within the fecal microbiota of all experimental groups (Anosim, permanova group significance p=.005, p=0.012). C. Differential expression sequencing (DESEQ) analysis of 16s bacterial rRNA. Pie charts represent the relative abundance of 11 significant bacterial genera within each experimental group with the following reported functional associations: Inflammation (Clostridia and Bacteroides), Cognition & Neurodevelopment (Oscillobacter and Lachnospiracae), and SCFA Production (Blautia, Lachnospiraceae, and Bacteroides).

Neurocognition

Neurocognitive phenotypes were assessed in all experimental groups at 45-50 days post-TBI (50 DPI). Two neurocognitive tests were employed—the zero maze to evaluate anxiety-like behavior and open field testing to assess generalized locomotion and exploratory behavior (Fig 2A-F). Open field testing revealed that SCFA-treated TBI mice maintained-non-TBI exploratory behavior as compared to vehicle-treated TBI mice as determined by the percent time spent in the center of the open field (11.7 ± 3% time in the center vs. 15.0 ± 6%; p=0.002) as well as by total distance traveled within the field (1097 ± 455.7 cm vs. 1460.1 ± 346.4 cm, p=0.02) (Fig. 2B). The zero-maze test (Fig. 2G-L) demonstrated preservation of non-TBI anxiety-like behavior in SCFA-treated TBI mice as compared to vehicle-treated TBI mice as assessed by the total distance traveled in the open region of the maze (152.3 ± 101.8 cm vs. 147.5 ± 60.0 cm, p=0.006) (Fig. 2H). The observed reduction in time spent in open arms may suggest the preservation of murine-typical anxiety-like behavior rather than SCFA toxicity, given the lack of motor deficit. This finding warrants further investigation into SCFA effects on anxiety-like behaviors in non-injured models.

Figure 2. Post-injury SCFA treatment alters TBI-induced cognitive deficits.

A. Graphical open field (OF) test representation. B. Quantification of test 45 days post TBI. C-F. A representative pattern of activity from each group. G. Graphical representation of elevated zero maze (ZM) test. H. Quantification of ZM test 45 dpi. I-L. A representative pattern of ZM activity from each group. For all analyses, the threshold for statistical significance was set at P < 0.05. ((p ≤ 0.05(*), 0.001(**), & 0.0001(***)). Data is shown as the mean ± SEM. 2-way ANOVA with Tukey's multiple comparisons test with n = 8/group.

Neuroimaging

Neuroanatomy was assessed using in-vivo, 3D, contrast-enhanced magnetic resonance imaging (MRI; Fig 3A) at 50 dpi. Fractional anisotropy (FA) maps, reflecting water diffusivity and serving as a surrogate for white matter connectivity, were generated from MRI data as we previously published ⁴⁵ (Fig. 3). These measurements serve as a direct measure of cortical volume loss over time. Figure 3A shows T2-weighted ventricle volume comparisons, representing cerebral spinal fluid (CSF) replacement of cortical volume loss as we have previously published⁴⁵. In untreated TBI mice, significant ventriculomegaly is observed, consistent with expected neurodegenerative outcomes and cortical volume loss as CSF replaces tissue. T2-weighted coronal sections (Fig 3A) and the comparison of ventricle size to whole brain volume between groups (Fig 3B) illustrate marked attenuation of ventriculomegaly in SCFA-treated mice versus untreated (p < 0.0001). Furthermore, FA mapping demonstrated significant preservation of white matter connectivity in SCFA-treated groups (Fig 3B; p <0.0001).

Figure 3. SCFA treatment attenuates cortical volume loss and preserves white matter connectivity after TBI.

A. Representative T2-weighted MR images acquired 50 days post-TBI (n=3 mice/group). Ventriculomegaly is highlighted with red arrows. B. Mean ventricle volume normalized to total brain volume demonstrates a reduction in injury severity with SCFA treatment (3.7%) compared to untreated TBI animals (6.1%) (p < 0.0001). C. A visual representation of whole-brain fractional anisotropy (FA) maps and (D) their corresponding quantification, derived from MRI data, reveal both injury (p < 0.0001) and treatment (p = 0.05) effects. Specifically, FA increased from 55.44 mm³ in untreated TBI mice to 72 mm³ in SCFA-treated TBI mice. Data are presented as mean ± SEM. For all analyses, the threshold for statistical significance was set at P < 0.05. ((p ≤ 0.05(*), 0.001(**), & 0.0001(***)). Statistical analysis was performed using a two-way ANOVA with Tukey’s multiple comparisons test (n = 4–5 per group).

FACS and Single-cell RNA sequencing

To further decipher potential molecular mechanisms of SCFA-mediated neuroprotection after TBI, CD45⁺ CD11b⁺ cells were immunophenotyped via flow cytometry at 30 days post-injury with two randomly chosen biologic replicates per group to control for individual biologic variability (Fig. 4). In total, 13,713 cells were sequenced from eight mice: 3,301 cells from mice undergoing sham injury followed by vehicle treatment, 1,206 cells from vehicle-treated TBI mice, 4,112 cells from SCFA-treated sham injury mice, and 5,094 cells from SCFA-treated TBI mice. After quality control regression of 8-11% of cells, 12,375 cells were used for integration in total (Table 1).

Figure 4. Post-injury SCFA treatment alters immune cell composition in TBI.

A. UMAP plot demonstrating cluster distribution of CD45+ cells across groups: 3,077 cells from vehicle-treated sham (SHAM), 1,046 cells from vehicle-treated TBI (TBI), 3,704 cells from SCFA-treated sham (SCFA), and 4,548 cells from SCFA-treated TBI (SCFA_TBI). B. Proportion of identified cell types across samples. C. Dot plot showing the scaled expression of canonical markers for each cluster, colored by the average expression of each gene across all clusters. Dot size represents the percentage of cells in each cluster with more than one read of the corresponding gene.

Table 1. The Proportion of identified cell types across samples in scRNA-seq.

ncMonocytes denote non-classical monocytes.

	SHAM	TBI	SCFA	SCFA_TBI
Microglia	84.82%	71.89%	81.72%	68.56%
Macrophage	3.64%	4.78%	3.10%	4.93%
Monocyte	0.81%	2.10%	2.32%	2.18%
ncMonocyte	1.43%	4.02%	1.78%	1.43%
Neutrophil	1.04%	1.63%	1.00%	1.45%
NK	0.81%	1.43%	0.59%	1.34%
T-cell	3.18%	6.98%	2.97%	10.38%
B-cell	3.64%	4.78%	3.10%	4.93%

While a lower input of cells was noted in vehicle-treated TBI mice, as we have seen previously ^43,54,55, visualization using uniform manifold approximation and projection (UMAP) showed uniform clustering across all groups (Fig 4A). Primary innate immune cells, including microglia, classical and non-classical monocytes, dendritic cells (DC), and macrophages, were identified using canonical immune markers. Adaptive immune cells were also seen, identifying natural killer (NK), T, and B cells (Fig. 4B). After annotating clusters using canonical immune cell markers (Fig 4C), proportional changes were compared between each group. This technique was employed to account for lower immune cell counts in the TBI group that could impact interpretation. While the decline of microglia and the increased infiltration of adaptive immune cells following TBI is well documented, we observed a marked increase of NK and NKT cells in all injured groups and an increase in T-cell involvement with SCFA treatment 30 days post-injury (Table 1) ^25,43,54.

Activated microglia are one of the primary drivers for chronic, low-level neuroinflammation following TBI. Therefore, transcriptional alterations within microglia clusters were examined across all experimental groups (Fig 5A-F). Homeostatic microglia markers were expressed across most cells, indicating a baseline presence even amid varying degrees of injury. Notably, disease-associated microglia (DAM) features were highly expressed in clusters adjacent to those with high inflammation features, a pattern consistent with typical responses to brain injury. Heat-shock response genes were notably expressed in cells with minimal-to-no inflammatory genes, suggesting a protective or recovery mechanism in some microglia. This expression pattern was consistent across the experimental groups (Fig. 6B), with injury-related cells clustering closer to cells with high DAM and inflammation features, as further detailed in (Fig. 5C).

Figure 5. Differential expression analysis of microglia based on injury and treatment.

A. Microglia feature scoring based on select gene expression. B. UMAP plot demonstrating clustering was obtained for the microglia population, annotated by their sample origins. C. Dot plot showing the scaled expression of select features from (A) for each experimental group, colored by the average expression of each gene. Dot size represents the percentage of cells expressing feature genes. D-E. Volcano plots illustrating the most significantly altered genes based on the Wilcoxon Rank Sum test and Benjamini-Hochberg correction (y-axis, adjusted p-values). The dotted line represents the adjusted p-value threshold of 10⁻⁵. D. Comparison between SHAM (vehicle-treated Sham injury) vs TBI (vehicle-treated TBI). E. Comparison between SCFA_TBI (SCFA-treated TBI injury) vs TBI (vehicle-treated TBI). F. Top 10 gene ontology terms from upregulated genes expressed in SCFA-treated TBI compared to vehicle-treated TBI microglia.

Most notably, SCFA treatment in the context of TBI led to increased expression of homeostatic microglia genes compared to microglia from vehicle-treated TBI mice. This finding coincides with our hypothesis that SCFAs are neuroprotective after TBI. Furthermore, there was a markedly greater heat shock transcriptional response (Hspa1a, Hspa1b, and Hspb1) in SCFA-treated TBI mice as compared to vehicle-treated TBI mice (Fig. 5E). Coincidentally, we observed a similar upregulation of heat shock transcripts in our previously published data using fecal microbiota transplant after TBI (Fig. 5E)⁸. However, minimal changes were observed in disease-associated microglia (Lpl, Cst7, Apoe, Lgals3, Lyz2, Mif, and Cd74), cytokine-associated (Ccl3, Ccl4, ll1a, and Tnf) and inflammatory genes (Ccl3, Ccl4, and Tnf) between SCFA and TBI groups. Interestingly, even in the absence of injury, SCFA supplementation reduced the expression of cytokine- and inflammation-associated genes in sham animals. The importance of this finding in both our previous work and the current study lies in numerous reports on the neuroprotective effects of the heat shock response.

To identify specific genes contributing to each experimental group, a list of differentially expressed genes between groups was generated (Sham vs. TBI, SCFA vs Sham, and SCFA_TBI vs TBI) (Fig. 5A-C). In response to TBI alone, Uba52 (encoding ubiquitin A-52 residue ribosomal protein fusion product 1) was the most significantly upregulated compared to sham, suggesting a potential upregulation of ubiquitin-mediated protein degradation pathways in response to injury (Fig 5D). Consistent with previously published data, Apoe (encoding apolipoprotein E) also showed a substantial increase in expression, indicative of heightened lipid metabolism and neuroinflammation post-TBI. Conversely, the heat shock protein genes, such as Hspa1a/b, Hsp90aa/b1, and Hspa8 (respectively encoding heat shock protein family A (Hsp70) member 1A/B, heat shock protein 90 alpha family class A/B member, and heat shock protein family A (Hsp70) member 8) demonstrated significant downregulation, indicating suppression of cellular stress responses in TBI-associated microglia. Next, microglia from SCFA_TBI mice were compared to microglia from vehicle-treated TBI to reveal the direct effects of treatment on gene expression (Fig 5E). Strikingly, genes that were upregulated in the TBI group, such as Uba52 and Gapdh, were significantly downregulated with SCFA supplementation. Furthermore, genes downregulated in response to TBI, such as heat shock protein genes Hsp90ab1 and Hspe1, were significantly increased with SCFA supplementation. This data suggests that 1) post-injury protein turnover may be a critical, keystone development in injury and SCFA modulation of injury and 2) a possible SCFA-mediated enhancement of the molecular chaperone response to TBI ^56,57. These gene expression changes underscore the complex molecular landscape influenced by SCFA treatment in the context of TBI.

After identifying the most significantly altered genes after SCFA treatment in TBI, a gene ontology (GO) analysis on upregulated genes within the microglia of SCFA-treated TBI mice was compared to the microglia of vehicle-treated TBI mice (Fig 5F). SCFA supplementation altered GO terms associated with immune response and homeostatic microglia functions such as synaptic pruning and vesicular activity. These findings suggest that TBI and SCFA treatment distinctly affect cellular pathways, potentially influencing neuroinflammation and recovery mechanisms post-TBI.

Discussion

The current study demonstrated that post-TBI administration of three SCFAs (acetate, butyrate, and propionate) significantly improved neurocognitive and neuroanatomic outcomes after a severe traumatic brain injury. This observed neuroprotection is certainly multifactorial. First, we link part of this benefit to SCFA-mediated stabilization of the gut microbial community structure, underscoring how bacterial metabolites influence both host health and the resilience of the gut microbiota itself (Fig. 1). This is notable because SCFAs, the main byproducts of bacterial fermentation of non-digestible carbohydrates, act as natural ligands for receptors on gut, immune, and neuronal cells, establishing a key bacteria-to-host signaling pathway⁵¹. Specific bacterial taxa—such as Lactobacillaceae, Ruminococcaceae, and Lachnospiraceae—are critical for producing these metabolites in sufficient quantities⁵⁰. Prior work, including our own studies, has shown a persistent loss of these fiber-fermenting bacteria after severe TBI, which correlates with increased brain ventricular volume and cognitive decline ^16,17,58. Our earlier fecal microbiota transplant (FMT) study reinforced this, showing that reestablishing a healthy gut microbiome after TBI reduced cortical volume loss and preserved cognitive function long-term ^16,17. Together with the current findings, these data highlight that gut taxa-derived metabolites play a vital role in long-term TBI pathophysiology.

The gut microbiota generates three primary SCFAs (acetate, butyrate, and propionate) and several secondary ones, such as formate, valerate, and caproate, some of which may be derived from the primary group⁵⁹. Here, four weeks of continuous SCFA supplementation stabilized the gut microbial community without the need for exogenous bacteria or additional dietary fiber (Fig 1). This protective effect likely reflects microbial cross-feeding, where fiber-fermenting bacteria share metabolites with other microbes, enhancing community resilience⁶⁰. Notably, SCFAs also have mild antibiotic effects on non-commensal microbes, which may prevent opportunistic overgrowth in treated TBI mice⁶¹. To our knowledge, this is the first report of cross-feeding benefits being harnessed under conditions of neurologic injury.

Beyond the gut, SCFA treatment provided robust long-term neuroprotection after just four weeks of supplementation (Fig. 2). MRI analyses showed that this intervention attenuated cortical volume loss and preserved white matter connectivity, aligning with improved neurocognitive outcomes (Figs. 2 & 3). The mechanisms are likely complex; however, our single-cell RNA sequencing revealed that SCFA supplementation shifted the composition of infiltrating immune cells in the injured brain. While immune cell infiltration can both exacerbate acute injury and support repair, pro-inflammatory profiles in microglia are known to worsen neuropathologic outcomes ^10,45. Our data recapitulate this, showing a persistent pro-inflammatory transcriptional profile in microglia one-month post-TBI. SCFA treatment mitigated these changes, producing an anti-inflammatory microglial profile marked by increased expression of neuroprotective heat shock proteins and reduced expression of neurodegeneration-related genes such as APOE and UBA52 (Fig. 4)^62,63. Concurrently, SCFA supplementation after TBI mitigated these transcriptional changes and resulted in an overall anti-inflammatory microglia transcriptional profile expressing log-fold increases in the neuroprotective heat shock genes with SCFA treatment after TBI (Fig. 4). Although SCFAs’ full immunomodulatory potential remains to be defined, these findings are consistent with our recent work linking SCFA-mediated T-cell activation to behavioral changes in an aged TBI model.^43,54,55

Our SCFA and prior FMT studies used consistent age, sex, weight, and injury parameters, resulting in comparable neuroanatomic and cognitive benefits (Fig. 5) 16. While further study is indicated, this suggests that both approaches may act through similar brain-gut axis mechanisms that influence peripheral immune infiltration and microglial states. However, FMT achieved even greater reductions in cortical volume loss and better preservation of white matter, implying that direct restoration of a healthy microbiome may provide additional benefits beyond metabolite supplementation alone.

The specific mechanisms by which short-chain fatty acids (SCFAs) modulate immune responses and contribute to injury tolerance within the host-associated microbiome are not fully understood. However, SCFAs, particularly propionate and butyrate, can diminish the activity of macrophages in the gut lamina propria involving nitric oxide, interleukins 6, and 12p40 pathways that are not typically associated with free fatty acid receptor signaling ^64,65. In the same work, SCFAs were shown to suppress the production of cytokines like IL-6 and IL-12p40 in mature human dendritic cells. These results from isolated cells were supported in vitro with acetate, revealing a modulatory effect on the release of inflammatory cytokines by primary microglia and astrocytes ^66,67. Histone deacetylases represent another burgeoning mechanism that can be considered in conjunction with our findings, as SCFAs have been shown to inhibit HDACs in brain microglia ^68,69. Though more study is required, these epigenetic changes to gene expression are thought to be one of the mechanisms by which SCFAs regulate microglial activity. Collectively, these findings support an essential role for SCFAs in modulating local and systemic immune functions, potentially influencing host metabolism and inflammatory pathways while mitigating gut dysbiosis linked to TBI.

Despite these promising findings, we recognize key limitations. Only male mice were used in this study, which restricts generalizability and may overlook important sex-specific differences in microbiome-TBI interactions. Female cohorts have since been incorporated into our ongoing work to address this gap. Additionally, we did not directly measure serum or fecal SCFA levels or calculate individual SCFA intake, as mice received SCFAs in drinking water ad libitum. This approach is consistent with cited studies but we acknowledge that precise intake data would enhance reproducibility and should be prioritized in future work. We also note that the scRNA-seq analyses were based on pooled brains from two mice per group. While this limits inter-animal comparisons, the effective sample size in scRNA-seq is defined by the thousands of individual cells profiled, which remains a robust exploratory approach for uncovering cell-specific transcriptional states. Utilizing pooled samples limits overall biologic variability. Nonetheless, future studies should expand this with larger, sex-balanced cohorts.

Lastly, while our behavioral analyses indicate preserved locomotion and exploratory behavior in SCFA-treated TBI mice compared to untreated TBI mice, we note that our measures did not include additional motor testing beyond total distance traveled in the open field and elevated zero maze. This is now acknowledged to ensure appropriate interpretation of these results.

In summary, our data provide compelling evidence that SCFAs—key metabolites of commensal gut bacteria—can modulate neuroinflammation and neurodegeneration after TBI in a clinically relevant treatment paradigm. By altering microglial transcriptional profiles and preserving gut microbial health, SCFAs show promise as a safe, effective, and potentially translatable intervention for human TBI. Future studies will refine dose-response relationships, expand sex-specific analyses, and further dissect the molecular pathways that connect the gut microbiome to the injured brain.

Data Availability

The corresponding author's data supporting this manuscript's findings are available upon request. The 16S rRNA gene and RNA sequencing data have been deposited to the NCBI BioProject data repository with the following dataset identifier: SUB14453560