Hyperbaric Oxygen Therapy Reduces the Traumatic Brain Injury–Mediated Neuroinflammation Through Enrichment of Prevotella Copri in the Gut of Male Rats

The animals were anesthetized by intraperitoneal injection of a mixture of ketamine (IM, 44 mg/Kg, i.m.; Nam Kwong Pharmaceutical, Taiwan), atropine (0.0633 mg/kg, i.m.; Suntong Chemical Industry Co., Ltd., Taiwan), and xylazine (6.77 mg/kg, i.m.; Bayer, Germany). Craniectomy with a radius of 2 mm was performed on a stereotaxic frame at 4 mm from the bregma and 3 mm from the sagittal suture of the right parietal cortex. After removal of the skull and implantation of the injury cannula, a fluid percussion device (VCU Biomedical Engineering, Richmond, VA) was attached to the animal via a luer lock connector, and the brain was injured with an impact force of 2.0 atm and 25 ms. This produced moderate brain trauma, as originally described by Mclntosh et al. [23]. Transient hypertensive responses, apnea, and seizures were observed immediately after fluid percussion injury, and they were used as criteria for successful TBI insult in rats. After the impact on the experimental rats, their tongues were immediately pulled out of the mouth to ensure the patency of the airway. There were no additional fatalities in the entire experimental setup because of this procedure.

Under ketamine anesthesia, the right femoral artery of rats was cannulated with a polyethylene tube (PE50) for blood pressure monitoring. Colon temperature was measured using an electronic thermometer (Model 43TE; YSI Corporation, Yellow Springs, OH) and a temperature probe (400 series; YSI Corporation). Core temperature was maintained within the range of 36–37 °C using lamp insulation. All recordings were conducted on a four-channel Gould polygraph.

Animals were randomly numbered and assigned to different study groups using a random number table. The animals were then allocated to three groups (n = 12 in each group): sham surgery plus normobaric air (NBA) (21% oxygen at 1 ATA), TBI (2.0 atm) plus NBA (21% oxygen at 1 ATA), and TBI (2.0 atm) plus HBO (100% oxygen at 2.0 ATA). Animals designated for HBO treatment (100% oxygen at 2.0 ATA) underwent a 60-min treatment immediately after TBI, as well as at 24 h and 48 h after post-TBI. Animals in the NBA group were exposed to room air, whereas those in the HBO group were placed in the HBO chamber. The chamber was flushed with 100% oxygen at a rate of 5 l/min to prevent CO accumulation. Decompression was carried conducted at a rate of 0.2 kg/cm²/min. The temperature of the pressure chamber was maintained between 22 and 25 °C.

Detailed information on the numbers of animals and results analysis is provided in Table 1.

The staining procedure for triphenyltetrazolium chloride (TTC) was conducted following the protocol described elsewhere. All animals were euthanized on the 3rd day after fluid percussion injury. Under deep anesthesia with ketamine, animals were intracardially perfused with saline. The brain tissue was then removed, immersed in cold saline for 5 min, and sliced into 1-mm sections. These brain sections were incubated in a 2% TTC solution dissolved in phosphate-buffered saline at 37 °C for 30 min and sequently transferred to a 10% formaldehyde solution for fixation. The injury volume, indicated by negative staining for TTC (reflecting dehydrogenase-deficient tissue), was assessed for each section using computer planimetry (computer-based image tool software) and summarized. The injury volume was calculated as 1 mm (thickness of the slice) multiplied by the sum of injury areas (in square millimetersmm2) in all brain sections [24].

Apoptotic neuronal cells were identified by terminal deoxynucleotidyl transfer-ase-mediated biotin–deoxyuridine triphosphate nick-end labeling (TUNEL) [25] staining on day 3 after TBI. These procedures followed those described previously [26]. The quantity of immunofluorescence assay–-positive cells was determined corresponding to the peri-injury cortex (magnification × 400). The results were expressed as the mean number of positive cells in all five sections from each rat, using computerized planimetry (Image-Pro Plus, Media Cybernetics, Inc., Rockville, MD). The following anti-bodies were used in this study: monoclonal mouse anti-NeuN antibody (NeuN, Millpore, MAB377) at 1:100 dilution, followed by Alexa-Fluor, 568 anti-mouse (IgG) antibody (Invitrogen, A11031) at 1:400 dilution.

On day 3 after TBI, activated microglia were assessed by detecting OX42-positive cells using an immunofluorescence assay [27]. The number of OX42 cells in the sample was measured in each slice and summed using a computerized planar measurement (Image Tools software based on PC-Q4Q5). Monoclonal mouse anti-OX42 antibody (ab78457; Abcam, Boston, MA), anti-glial fibrillary acidic protein (GFAP) antibody (MA5-15,086; Thermo USA) and polyclonal goat anti-TNF-α antibody (sc-1351; Santa Cruz Biotechnology Inc, Santa Cruz, CA) was used in this study and used Alexa-Fluor, 568 antimouse (IgG) antibody (A11031; Life Technologies Co, Grand Island, NY), and DyLight 488 antigoat (IgG) antibody (ab96931; Abcam), respectively.

The genomic DNA was extracted from the fecal samples by using the QIAmp Fast DNA Stool Mini Kit (Qiagen, Germany) with the modified instructions. Briefly, the stool sample was centrifuged at 13,200 rpm for 10 min to remove the storage buffer and lysis with the InhibitEX buffer. Add the proteinase K and ethanol to obtain the processed su-pernatant. Finally, the supernatant was washed with the QIAamp spin column and eluted with the elution buffer. The concentration was assessed by NanoDrop 2000 and then performed × 10X dilution with elution buffer.

The gut microbiome library was constructed with the standard V3–V4 region of the 16S ribosomal RNA gene. PCR amplification was performed with KAPA HiFi hotstart readymix (Roche) and purified using AMPure XP magnetic beads (Beckman Coulter, Brea, CA). The amplified PCR products were assessed for quality using a Fragment Analyzer (Advanced Analytical, Wood County, WV) and quantified using a Qubit 3.0 Fluorometer. Subsequently, the library was sequenced on a MiSeq platform (Illumina, San Diego, CA) with paired-end reads (2 × 300 nt), generating at least 100,000 reads for each sample.

The raw paired‐end reads were be trimmed, and that pass the quality filters assigned to operational taxonomic units which ≥ 97% similarity with the Greengene Data-base (v13.8). Operational taxonomic units (relative abundance), alpha diversity, beta diversity (PCoA), and heatmap were performed with the MicrobiomeAnalyst (https://​www.​microbiomeanalys​t.​ca/​MicrobiomeAnalys​t/​upload/​OtuUploadView.​xhtml), CLC genomics workbench (Qiagen, Germany), and GraphPad Prism 8 (GraphPad Software, La Jolla, CA). The abundance bacteria analysis (Linear discriminant analysis Effect Size [LEfSe]) and functional analysis (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) were performed by Galaxy/Hutlab website (http://​huttenhower.​sph.​harvard.​edu/​galaxy/​).

All of the data were analyzed using SigmaPlot, version 10.0 for Windows (Systat Software, San Jose, CA) in this study. Results are expressed as the means ± standard deviations of the mean for the experiments. The t-test, Spearman’s correlation, one-way analysis of variance, and post hoc analysis were used to compare the group differences. A p value less than 0.05 was considered statistically significant. The SPSS program was used for the above calculations.

At 72 h after TBI, TTC-stained sections revealed a significant increase in brain injury volume in the NBO-treated TBI group (136 ± 13 cm³) compared with sham TBI controls (0 ± 0 cm³, ***p < 0.001, n = 6 in each group). However, TBI-induced brain injury volume was significantly reduced by HBO treatment (94 ± 8 cm³, p = 0.045, n = 6 in each group, see Fig. 2A). The neuroinflammation markers Brain OX42 + TNF-α (Fig. 2B) and Brain-GFAP + TNF-α (Fig. 2C) also exhibited a significant increase in the brain injury volume of the NBO-treated TBI group, which was effectively reversed by HBO treatment following TBI. Furthermore, Fig. 1D illustrates that TBI-induced neuronal apoptosis marker Brain Neu-N + TUNL was significantly mitigated by HBO treatment following TBI (n = 6 in each group). Importantly, there exists a statistically significant positive correlation between neuroinflammation, brain injury volume, and neuronal apoptosis (see Fig. 3).

On the 3rd day after moderate TBI treatment (TBI 2.0 atm), we used online tools, specifically alpha and beta diversity analysis via Microbiome Analyst, to assess the diversity of the microbiota. We found that the total species diversity (alpha diversity) showed no significant changes: (Sham VS TBI 2.0, p = 0.207; TBI 2.0 VS TBI 2.0 + HBO, p = 0.111; sham VS TBI 2.0 + HBO, p = 0.691) (see Fig. 4A). Additionally, the overall bacterial composition (beta diversity) did not exhibit significant differences among the groups (see Fig. 4B).

We further used online tools, including LEfSe and Galaxy HutLab, to analyze the compositions of the core gut microbiome. We observed 14 abundant bacteria in the sham vs. TBI 2.0 group (Fig. 5A), 8 abundant bacteria in the sham vs. TBI 2.0 + HBO group (Fig. 5B), and 16 abundant bacteria in the TBI 2.0 vs. TBI 2.0 + HBO group, n = 6 in each group (Fig. 5C). Additionally, we employed an online Random Forest tool (Random Forester Analysis, Microbiome Analyst) to rank the importance of variables in a regression or classification problem. Based on the Random Forester analysis, our results indicated the Collinsella aerofaciens and Prevotella copri might be the most important gut microbes in HBO treatment for brain trauma (Fig. 5D).

Based on the LEfSe analysis (Fig. 5A–C), we identified five abundant bacterialgenera (Deinococcus spp., Clostridium spp., Micrococcus spp., Parabacteroides spp., and Lactobacillus spp.) and two abundant bacterial species (Prevotella copri and Rothia nasimurium) among the three groups. Further analysis revealed that Deinococcus spp., Clostridium spp., and Prevotella copri were significantly decreased following brain trauma treatment (TBI 2.0) (n = 6 in each group). In contrast, HBO treatment (TBI 2.0 + HBO) significantly increased the abundance of Micrococcus spp. and Rothia nasimurium (Fig. 6).

We analyzed the correlation abundant bacteria and brain injury volume, neuroinflammation, and neuronal apoptosis following TBI and treatment with HBO using Spearman’s correlation analysis. Our results revealed that Prevotella copri and Deinococcus spp. exhibited a significant negative correlation with TTC stain, Brain-GFAP + TNFα, and Brain-OX42 + TNFα (Fig. 7A, B) (n = 6 in each group). Additionally, both Prevotella copri and Deinococcus spp. showed a positive correlation trend with neuronal apoptosis.

We used online tools (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States, Galaxy/HutLab) to analyze the effects of abundant bacteria on metabolic pathways among three groups. Our analysis revealed 39 functional pathways in the sham and TBI2.0 group (Fig. 8A), while only one functional pathway was identified in the sham and TBI2.0 + HBO group (Fig. 8B) (n = 6 in each group).

Furthermore, our results disclosed that TBI significantly decreased neuronal function in the glutamatergic synapse (p = 0.034), Galactose metabolism (p = 0.027), and sphingolipid metabolism pathways (p = 0.019). Additionally, TBI significantly modulated immune response pathways such as the NOD-like receptor signaling pathway (p = 0.007) and protein degradation pathways like the Proteasome (p = 0.007) (Fig. 9A–E) (n = 6 in each group). These trauma-induced effects might be reversed by HBO therapy.

We conducted Spearman’s correlation analysis to explore the correlation between neuroinflammation, neuronal apoptosis, and bacterial metabolic pathways following TBI treatment (see Fig. 10). Our findings revealed that Glutamatergic synapse (R = − 0.607, p = 0.011), Galactose metabolism (R = − 0.487, p = 0.049), and Sphingolipid metabolism (R =  − 0.549, p = 0.024) exhibited a significant negative correlation with neuronal apoptosis (Brain-NeuN + TUNEL). Both the NOD-like receptor signaling pathway (R = 0.475, p = 0.056) and proteasome (R = 0.475, p = 0.056) displayed a positive correlation trend with neuronal apoptosis. Additionally, we observed a positive correlation trend between both the NOD-like receptor signaling pathway (R = 0.475, p = 0.056) and the proteasome (R = 0.475, p = 0.056) and neuroinflammation (n = 6 in each group).

Finally, we conducted Spearman’s correlation analysis to examine the correlation of between abundant bacteria and their metabolic pathways (Fig. 11). Prevotella copri exhibited a significant negative correlation to with the NOD-like signaling pathway (R =  − 0.544, p = 0.032) and the proteasome (R =  − 0.546, p = 0.031)., However, in contrast, it showed a modest positive correlation with sphingolipid metabolism, specifically in neuronal function (R = 0.309, p = 0.244) (n = 6 in each group).

Consistent with previous studies, our results demonstrate that HBO has beneficial effects on TBI-induced neuroinflammation, apoptosis, and brain injury volume [8, 10]. Additionally, we present new findings indicating a significant positive correlation between neuroinflammation and brain damage volume or neuronal apoptosis (see Fig. 3). These findings suggest that neuroinflammation plays a crucial role in the injury process following TBI and highlight the inhibition of inflammatory responses as a potential target for brain injury treatment.

When considering the gut bacterial composition on the 3rd day after TBI, our findings on regarding alpha diversity revealed a nonsignificant increase after TBI (see Fig. 4). Interestingly, this result contrasts with significant gut microbiota depletion observed in other animal models after TBI [22, 28, 29]. We hypothesize that this discrepancy may be attributed to differences in the nature of the impact injuries. For instance, our study used fluid impacts with a force of 2.0 atm, whereas other studies employed direct impacts using a pneumatic impactor (5.0 m/s; 250 μs dwell time and 2 mm depth), or impacts induced by placing a 50 g load at a height of 10 cm and loosely applying the impacts along the guide rails [28, 29]. These varying impact mechanisms may result in differential effects on the intestinal flora. Therefore, further investigations are warranted to elucidate whether HBO therapy exerts distinct effects on gut microbiota composition in response to different levels of brain injury.

Evidence has shown that TBI could alter some gut bacterial taxa. Treangen et al. [30] demonstrated that there was an early decrease in genus-level or species-level abundance of bacterial gut microbiome 24 h after the experimental TBI model. Rogers et al. [31] have shown early depletion of commensal bacteria and enrichment of potential pathogens at 72 h after TBI in pediatric severe TBI. In the current study, we found that five genus-level and two species-level bacteria of core gut bacteria decreased at 72 hours after TBI. These TBI-induced bacterial depletion effects could be further altered is determined by whether the bacteria themselves are aerobic or anaerobic through HBO therapy (Fig. 5). These results are consistent with HBO therapy being able to alter the composition of the gut microbiota [32].

Being as anaerobic bacteria, the abundance of Parabacteroides spp. and Lactobacillus spp. were significantly decreased after HBO therapy. In contrast, as aerobic bacteria, the abundance of Micrococcus spp. and Rothia nasimurium increased. This result is consistent with Wu et al. [33] report that increased tissue oxygenation could directly influence microbes, such as reducing Anaerostipes. Because 90% of the intestinal tract is composed of anaerobic bacteria [34], HBO therapy should be helpful in the development of pathologies caused by the proliferation of anaerobic bacteria.

An increase or decrease in Prevotella copri levels compared to normal values will influence disease progression. Increased levels of Prevotella copri have been reported in patients with new-onset untreated rheumatoid arthritis [35] and irritable bowel syndrome [36], while decreased levels have been detected in patients with neurodegenerative diseases [37]. Consistent to with findings in neurodegenerative diseases, our current study revealed a significant decrease in Prevotella copri levels after TBI, which could be reversed after HBO therapy (Fig. 6). Additionally, its level was found to be significantly negatively correlated with neuroinflammation and brain injury volume (Fig. 7). These results suggest that using HBO therapy to repair gut dysbiosis caused by TBI may represent a novel mechanism for reducing inflammation in the brain.

The glutamatergic synaptic pathway is interconnected with numerous other neurotransmitter pathways and is pivotal in a wide array of normal physiological functions [38]. A delicate equilibrium between sphingolipid synthesis and degradation is typically essential for various biological processes [39, 40], and alterations in their metabolism could significantly impact brain homeostasis and function. Nonetheless, the detailed mechanism requires further clarification in future studies.

NOD-like receptor has been reported to Nod1 as novel regulators for the stress response, 5-HT biosynthesis, and signaling. These results further indicate that Nod1 may contribute to a previously unrecognized signaling pathway in the gut-brain axis [41]. NOD-like receptor C5 promotes positively regulate neuroinflammation, suppresses neuronal survival and dopaminergic degeneration in Parkinson’s disease (PD), and may serve as a marker of glial activation [42]. In this study, we identified a positive correlation trend between the NOD-like receptor signaling pathway and the proteasome with neuronal apoptosis and neuroinflammation. Finally, we observed a significant negative correlation between Prevotella copri levels and the NOD-like signaling pathway and proteasome activity. Therefore, we suggest that attenuating TBI-induced neuroinflammation may be related to the inhibition of NOD-like receptor signaling and proteasome pathways, and may also be part of the neuroprotective mechanism of HBO. However, this speculation needs to be further clarified in the future.

There are some limitations to our study. Firstly, it is important to address the width of the therapeutic window for HBO treatment. In the current study, the clinical significance of initiating HBO treatment immediately after impact was found to be limited. In future studies, we plan to delay the first HBO exposure until 3, 6, 12, or 24 h after TBI to enhance clinical relevance and comprehensively assess the effect of HBO on TBI. Second, we employed three groups to assess the effects of HBO on TBI in rats, adhering to our established study design from previous research. there were only three groups of animals: sham + normal oxygen, TBI + normal oxygen and TBI + hyperbaric oxygen. It would be more beneficial to compare normobaric 100% oxygen to HBO at 2 ATA and normobaric room air at 21% oxygen. This comparison would help determine whether there is any additional benefit from HBO compared to the interventions clinicians can attempt without placing the patient in an HBO chamber. In addition, without a control group with a sham procedure and exposure to hyperbaric oxygen, one cannot determine whether the change of gut bacteria is related to hyperbaric oxygen or the response to both TBI and hyperbaric oxygen. Third, we only examined the effect of short-term HBO interventions after injury, whereas a 3-day assessment is evident for assessing parameters related to neuroinflammation in secondary injury due to TBI [8]. Fourth, we used only moderate fluid impact injuries (2.0 atm) to assess the effects of HBO on the brain and gut. Further studies are needed to determine whether HBO has a differential effect on the severity of brain injury. Fifth, although OX42 has been identified as a marker for microglia, it is crucial to acknowledge that, in addition to microglia, OX42 also stains for granulocytes. Likewise, although TUNEL staining has been established as a marker for apoptotic cells, it may produce positive results in necrotic and autolytic cells, as well. Therefore, a more specific marker is considered essential for future studies. Sixth, we did not assess histopathologic changes in the gut although it is now known that gut damage occurs after TBI [15]. Finally, importantly, Prevotella Copri abundance was negatively correlated with NOD-like receptor signaling and proteasome pathways. However, the causal relationship between Prevotella copri and NOD-like receptor signaling and proteasome pathways remains unknown. In addition, it is unclear whether supplementation with HBO, Prevotella copri ameliorates TBI-mediated neuroinflammation and apoptosis. Therefore, more studies are needed to explore the above relationships and potential therapeutic effects.