Sepsis-Induced Gut Dysbiosis Mediates the Susceptibility to Sepsis-Associated Encephalopathy in Mice

The neurological scores and survival rates of mice at different time points after surgery were recorded. In the mice that underwent operation, the scores of the neurological assessment were lowest at 2 h; this was followed by a gradual recovery of the neurological responses between 2 and 24 h. The neural reflexes of sham-operated mice (CON) returned to normal within 24 to 48 h. However, the scores of the cecal ligation and puncture (CLP)-operated mice were decreased at 36 h, which was followed by a gradual increase from 48 h to 7 days (Fig. 1A). In addition, the standard deviation of scores in the CLP group showed an obvious increase at 36 h (Fig. 1A), which indicates considerable interindividual variability in neural reflexes of septic mice at 36 h. Survival analysis showed high mortality in CLP-operated mice within 48 h (Fig. 1A). Therefore, mice with a neurological score of ≤6 at 36 h were defined as SAE susceptible (SES), while mice with a score >6 were defined as SAE resistant (SER). There was no significant difference in the neurological scores between the SES and SER groups in the first 24 h; however, the scores in the SES group were lower than those in the SER group at 36 h and thereafter (Fig. 1B). In addition, the survival of SER mice was better than that of SES mice (Fig. 1B).

We further compared the extent of behavioral changes among these three groups. Compared to the control group, mice in the CLP group performed worse in the open field test (Fig. 1C). The total distance traveled, duration spent in the center, and ratio of center distance to total distance of mice in the SES group were significantly lower than those in the SER group at 36 h. These findings suggest that the SES mice showed greater anxiety and worse mobility (Fig. 1C). In addition, we evaluated systemic inflammation and neuroinflammation in the mice that underwent an operation. Interleukin-β (IL-β) and tumor necrosis factor alpha (TNF-α) levels in both the serum and cortex of CLP-operated mice were significantly elevated compared to those in the sham-operated mice. In both the serum and cortex, SES mice showed higher levels of IL-β and TNF-α than SER mice, which indicates more pronounced systemic inflammation and neuroinflammation in SES mice (Fig. 1D).

To explore whether the susceptibility of mice to SAE is related to preoperative differences in the microbiota or to sepsis-induced differences in the microbiota, we analyzed the gut microbiota composition of mice before and after CLP. The Shannon, Simpson, Ace, and Chao indices were calculated to evaluate the bacterial α-diversity in various groups, and principal-component analysis (PCA) was used to assess β-diversity. There was no significant difference between SES and SER mice with respect to α- or β-diversity before CLP (see Fig. S1A to C in the supplemental material), suggesting that preoperative gut microbiota may not be a factor influencing susceptibility to SAE. However, the gut microbiota of the mice was markedly disrupted after the CLP operation (see Fig. S2A to C in the supplemental material). Compared with control mice, the richness and diversity of the intestinal microbiota were significantly decreased in SES mice, whereas SER mice had mildly dysregulated intestinal microbiota. Specifically, SES mice showed a lower Shannon index and a higher Simpson index than SER mice (Fig. 2A), suggesting a decrease in bacterial diversity in SES mice compared with that in SER mice. PCA data also showed distinctive microbial communities in the fecal microbiota between the SES and SER groups after operation (Fig. 2B and C). There was considerable interindividual variability in gut microbiota composition at the phylum level in mice that underwent an operation (Fig. 2D).

We further analyzed the differences in gut microbiota composition between the SES and SER groups. At the family level, Enterobacteriaceae enrichment was observed in SES mice (Fig. 2E). Moreover, linear discriminant analysis effect size (LEfSe) revealed significant differences between the two groups. In particular, the order Enterobacterales, family Enterobacteriaceae, phylum Proteobacteria, and class Gammaproteobacteria were more enriched in the SES group, while the class Bacteroidia, phylum Bacteroidota, and order Bacteroidales were enriched in SER mice (Fig. 2F and G).

After identifying significant differences between SES mice and SER mice with respect to postoperative gut microbiome composition, we sought to determine whether these differences contributed to the differential susceptibility to SAE. We performed an FMT experiment (Fig. 3A). After antibiotic treatment to deplete the intestinal microbiota for 5 days, mice received feces from healthy (CON-FMT), SES (SES-FMT), and SER (SER-FMT) mice (Fig. 3A). All mice that received feces from SES mice died within 48 h. The survival rate was 40% for mice receiving feces from healthy mice and 30% for mice receiving feces from SER mice (Fig. 3B). These results suggest that the gut microbiota of SER mice could provide a survival advantage after CLP compared to SES mice. In addition, mice in the CON-FMT group showed the lowest levels of proinflammatory cytokines in the serum and cortex among these three groups. In contrast to mice gavaged with feces from SER mice, mice that received feces from SES mice had significantly higher levels of IL-1β and TNF-α both in the serum and cerebral cortex (Fig. 3C). These data indicate that CLP-induced neuroinflammation could be transferrable by gut microbiota and that the gut microbiota from SER mice was a key factor in the resistance of mice to SAE.

To explore the sepsis-induced functional differences of the gut microbiota between SES and SER mice, PICRUSt analysis, a computational approach to predict microbial physiological function at the genomic level, was conducted. Significant differences in the expression level of several genes associated with metabolism were observed. More specifically, the genomic abundance of certain pathways, such as the lipid metabolism, carbohydrate metabolism, terpenoid, and polyketide metabolism, was significantly diminished in SES feces compared to SER feces (Fig. 4A). Nontargeted metabolomics analysis of the gut microbiota was performed to further explore metabolic functions in SES and SER mice. PCA showed separation of the metabolite profiles of each group (Fig. 4B). Additionally, some metabolites were differentially enriched in the feces of the two groups (Fig. 4C). Indole-3-propionic acid (IPA) is a metabolite that was found to be neuroprotective in previous studies (11). IPA was significantly enriched in the feces of SER mice, implying that it may also play a protective role in the pathogenesis of SAE (Fig. 4C). Moreover, the serum levels of IPA in SER mice were higher than those in SES mice (Fig. 4D).

Since serum IPA concentrations differed between SER and SES mice, we wondered whether IPA concentrations in mouse serum were affected by the intestinal microbiota. Mice were administered antibiotics (ABX) for 5 days to deplete their intestinal microbiota, followed by gavage with feces from healthy mice for 3 days to restore their intestinal microbiota (Fig. 4E). The results showed that the IPA levels in serum decreased significantly after ABX treatment, but the IPA levels partially reverted after administration of feces from healthy mice (Fig. 3F). Our data clearly demonstrate that changes in the gut microbiota affect serum IPA levels.

Next, we treated mice with IPA or saline to assess the protective effect of IPA against CLP-induced SAE and death. Survival analysis showed better survival for IPA-treated mice than saline-treated mice (Fig. 5A). Additionally, neurologic assessment showed greater activity in the neurological reflexes of IPA-treated mice at 36, 48, and 72 h compared to that of saline-treated mice subjected to CLP (Fig. 5B). Open field tests showed a significant increase in the duration spent in the center and the ratio of the center distance to the total distance of IPA-treated mice compared with that of saline-treated mice. However, no significant between-group difference was observed with respect to the total distance traveled. These findings suggested that IPA reduced anxiety but did not improve mobility in CLP-operated mice (Fig. 5C). The Morris water maze test showed that the frequency of crossing the original platform location and duration spent in the platform quadrant during the probe trial on the sixth day in the IPA-treated group were markedly increased compared with those of the saline-treated group. However, there was no significant difference in the swimming speeds (Fig. 5D). These data indicate that IPA may have restored the CLP-induced cognitive impairment in the CLP model. In addition, IPA-treated septic mice had lower levels of IL-1β in the serum and cortex than saline-treated mice (Fig. 5E and F). There was a high degree of overlap between IL-1β-positive cells and microglia (Fig. 5F), suggesting the involvement of microglia in the pathophysiology of SAE development.

Based on the results of immunofluorescent staining, we speculated that the levels of IL-1β in the cortex of CLP mice may be reduced by IPA and that microglia were the main effector cells involved in this pathophysiology. Therefore, we wondered whether IPA suppresses cortical inflammation by inhibiting the activation of the NLRP3 inflammasome in microglial cells. The NLRP3 inflammasome has been reported to be a key player in the development of SAE (19), but the effect of IPA on the NLRP3 inflammasome is unclear. Western blotting analyses showed an apparent increase in the levels of NLRP3, pro-IL-1β, and IL-1β in the cortex of CLP-operated mice compared with those of sham-operated mice. These findings suggest a critical role of the NLRP3 inflammasome in SAE. Treatment with IPA resulted in a decrease in the levels of NLRP3 and IL-1β in the cortex compared with those in CLP mice treated with saline. In addition, mice in the IPA and saline groups showed similar levels of procaspase-1 and pro-IL-1β in the cortex (Fig. 6A). These results suggest that IPA could inhibit NLRP3 expression and IL-1β production.

Sufficient NLRP3 protein expression is essential for NLRP3 inflammasome formation and activation (20). Given that IPA negatively regulates NLRP3 expression, we tested whether IPA could inhibit NLRP3 inflammasome activation in vitro. NLRP3 activation promotes the cleavage of procaspase-1, and then mature caspase-1 promotes the cleavage of pro-IL-1β to IL-1β (20). Primary microglia were pretreated with IPA for 1 h, followed by stimulation with lipopolysaccharide (LPS) for 12 h. After stimulation with LPS, the levels of NLRP3, cleaved caspase-1, and IL-1β in cells showed a significant increase compared with controls. However, the increase was markedly restored after an LPS challenge in the presence of IPA (1 mM) (Fig. 6B and C). These data indicate that IPA inhibits NLRP3 inflammasome activation and the subsequent IL-1β secretion.

The aryl hydrocarbon receptor (AhR) has been shown to negatively regulate NLRP3 inflammasome activity (21). AhR, a ligand-dependent transcription factor, can be activated by diverse ligands including indole-based compounds (22, 23). Thus, we hypothesized that IPA inhibited the activation of the NLRP3 inflammasome, and this inhibition could be reversed by an AhR antagonist. As shown in Fig. 6D to F, pretreatment with an AhR antagonist (CH223191↗) attenuated the effects of IPA on the inhibition of NLRP3 expression, caspase-1 cleavage, and IL-1β secretion in the primary microglial cells after an LPS challenge in vitro. These results suggest that the inhibitory effect of IPA on NLRP3 inflammasome activation and IL-1β secretion in microglia after LPS challenge might be mediated by AhR.

In summary, by associating a microbial metabolite, IPA, as a neuroprotective compound, our study suggests that the gut microbiota is an upstream regulator of SAE and that the variability in sepsis-induced gut dysbiosis mediates the differential susceptibility to SAE in CLP-induced experimental sepsis mice. Our findings reveal potential mechanisms by which the gut microbiota mediates SAE susceptibility and enrich our understanding of the “microbiota-gut-brain” axis during sepsis progression.

Specific pathogen-free male C57BL/6 mice (age, 6 to 8 weeks) were used in this study. All experimental animal studies were approved by the local Animal Care and Use Committee of Guangdong Provincial People's Hospital. A mouse model of sepsis was induced by performing CLP (9). Briefly, mice were anesthetized with pentobarbital. After skin preparation and disinfection of the abdomen, a midline incision of approximately 1 cm was made to expose the cecum. The middle of the cecum was ligated and a single through-and-through puncture was performed with a 21-gauge needle between the ligation site and end of the cecum. After extruding a small amount of fecal material through the puncture, the cecum was gently repositioned in the abdominal cavity and the incision was sutured. Sham-treated mice underwent laparotomy and manipulation to expose the cecum without ligation and puncture. All mice were resuscitated by subcutaneous injection of 1 mL of saline. Before modeling sepsis, mice in the indole-3-propionic acid (IPA) (Sigma; 220027) treatment group received IPA by gavage (25 mg/kg of body weight dissolved in saline) once daily for 5 days and once within 1 day after modeling (modified treatment protocol based on previous studies [29, 38] and our pilot trials). Equal amounts of saline were administered to mice in the control group. All mice were provided ad libitum access to feed and water and housed in a controlled environment. Mouse feces were collected 1 day before and 2 days after the CLP operation and stored at −80°C. Some mice were sacrificed after 36 h to harvest tissues.

FMT was performed according to the modified method described elsewhere (9). Briefly, male C57BL/6 mice (age, 6 to 8 weeks) received feces from donor mice (control, susceptible, and resistant) after depletion of the gut microbiota by antibiotic treatment. The antibiotic regimen that was administered consisted of 100 mg/kg vancomycin, 200 mg/kg neomycin sulfate, 200 mg/kg metronidazole, and 200 mg/kg ampicillin. The regimen was administered by gavage once daily for 5 days. The fecal administration regimen consisted of resuspending donor feces in saline at a concentration of 0.125 g/mL and then administering 0.15 mL of this suspension to mice by oral gavage once daily for 3 days. After 3 days, mice were subjected to CLP and some were euthanized 12 h later to harvest tissues.

Pure primary microglia cultures were obtained from the cortex of 1-day-old C57BL/6 mice according to a modified method described elsewhere (39). In brief, mouse cortices obtained by dissection under aseptic conditions were cut into pieces and digested with trypsin (final concentration, 0.125%) at 37°C for 10 min. After the addition of fetal bovine serum (FBS) to stop the reaction, the suspension was filtered through a 200 mesh filter and then centrifuged at 1,000 rpm for 5 min. Cells were resuspended in Dulbecco modified Eagle medium (DMEM)/F-12 (10% FBS, 1% penicillin/streptomycin), seeded in several flasks precoated with poly-l-lysine (PLL) (0.1 mg/mL), and grown at 37°C with 5% CO₂ for 12 to 14 days. Small amounts of medium were added every 3 days thereafter, and the growth of the cells was observed under a microscope. Culture flasks were shaken for 1 h (200 rpm, 37°C) to detach microglial cells, and the medium containing microglia was collected for centrifugation (1,000 rpm, 5 min). The supernatant was removed, and the obtained pure microglia were resuspended in DMEM/F-12 (10% FBS, 1% penicillin/streptomycin) and seeded onto subcultures precoated with PLL. After 24 h of pure microglial culture, the cells were treated with IPA (0.5 mM, IPAL group; 1.0 mM, IPAH group). Then, they were dissolved in dimethyl sulfoxide (DMSO) for 1 h followed by LPS (1 μg/mL) stimulation for 6 h. Cells in the LPS group were stimulated with LPS only, and cells in the control group were not stimulated. Cells in the CH223191↗ +IPAH group were pretreated with CH223191↗ (10 μM, dissolved in DMSO) for 1 h. Then, they were treated with IPA (1.0 mM) for 1 h followed by LPS (1 μg/mL) stimulation for 6 h.

Other materials and methods are described inin the supplemental material. Text S1

Data are presented as the mean ± standard deviation. Statistical analysis was conducted with either Student's t test or analysis of variance (ANOVA). Post-hoc analysis was performed for the ANOVA testing. P values of <0.05 were considered indicative of statistical significance.

The data sets used and analyzed during the current study are available from the authors upon reasonable request, and some have already been included in the paper. Raw sequence data of the microbiota that support the findings in our study have been deposited into NCBI’s Sequence Read Archive under accession number PRJNA779842↗.