Translocation of bacteria from the gut to the brain

We previously studied a functional connection between the gut microbiome and chronic liver disease in the multidrug resistance gene 2 knockout (Mdr2^−/−) mouse model of cholestatic liver disease [29]. Mdr2 is a homolog of human biliary transport protein MDR3, the absence of which leads to a progressive development of intrahepatic cholangitis, fibrosis, and cirrhosis in mice, resembling primary sclerosing cholangitis in humans. In Mdr2^−/− mice, we observed the disruption of bile flow (i.e., cholestasis) and altered composition of gut bacteria in the intestine, including that of Lactobacillus [29]. Here, we tested how diet alteration might impact the gut microbiome in this model by feeding Mdr2^−/− mice an atherogenic, high-fat, Paigen diet. Total bacterial colony-forming units (CFU) were comparable in fecal samples and the ileum of Mdr2^−/− mice fed regular diet (Control) or Paigen diet (Paigen). However, after 9 days of Paigen diet, Mdr2^−/− mice exhibited an enrichment of specific bacterial species in the gut such as Staphylococcus, Bacteroides, and Akkermansia, and a reduction of gut commensal lactobacilli in the feces, as compared with mice fed a Control diet (Fig 1A). Similarly, an increased relative abundance of the bacterial families Staphylococcaceae (0.5% to 15.4%), Bacteroidaceae (0% to 21.8%) and Akkermansiaceae (0.16% to 15%), and a decreased abundance of Lactobacillaceae (14.6% to 0.8%), were found in fecal samples from mice fed Paigen diet (S1 Fig). The alterations in gut microbiome composition correlated with increased gut barrier permeability in mice fed Paigen diet (Fig 1B). To determine whether the intestinal permeability in these Paigen diet-fed mice might lead to bacterial dissemination out of the intestine, we tested for levels of bacteria in fecal pellets and the ileum, as well as in organs such as the lung, heart, kidney, spleen, brain, and the blood (Figs 1C–1F and S2). While we did not observe bacteria in most systemic organs or the blood (sensitivity <5 CFU, S1 Table), we were surprised to observe small numbers of bacteria in the brains of Paigen diet-fed mice (Fig 1E). Due to the low numbers of bacteria detected in the brains of Paigen diet-fed mice, we used both quantitative (i.e., Mann–Whitney U test, MWU) and qualitative (i.e., Fisher’s Exact test, FE) analyses to determine statistical significance and a positive association between brains harboring bacteria in samples from the Paigen diet versus Control diet-fed mice (Fig 1E). Interestingly, these bacteria were not present in the meninges or cerebrospinal fluid (CSF), highlighting that this was not meningitis (S2E and S2F Fig). Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry identified the bacteria isolated from the brains as Staphylococcus species (Staphylococcus xylosus and Staphylococcus sciuri) and Enterococcus faecalis, consistent with the presence of staphylococci and enterococci in the fecal samples and ileum of Paigen diet-fed Mdr2^−/− mice (Figs 1G–1I and S3). The localization of a low level of bacteria in the brains of Paigen diet-fed mice was not due to increased blood–brain barrier permeability, which was not observed in these mice or mice fed the Control diet (Fig 1J). To prevent contamination during tissue processing, we used extremely careful steps in preparation and during sample collection (summarized in Box 1 and S4 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Bacteria translocate from the gut to the brain via vagus nerve.

Paigen diet was fed to 12-week-old Mdr2^−/− mice for 9 days. (A) Fecal pellets collected at day 9 were processed for 16S rRNA gene sequencing. Cladogram representation showing the phylogenetic relationship between bacterial families between Control diet (green) and Paigen diet (red) groups. (B) Intestinal barrier permeability in Control diet- and Paigen diet-fed mice. (C–F) Total colony-forming units (CFU) of bacteria per gram of (C) fecal pellet and (D) ileum, (E) per brain, and (F) per ml of blood. (G–I) CFU of Staphylococcus xylosus per gram of (G) fecal pellet, (H) ileum, and (I) per brain. (J) Brain permeability index of Control diet- and Paigen diet-fed mice. (K) Total CFU of bacteria per ~2 cm of vagus nerve. (L) Total CFU/brain following unilateral right cervical vagotomy of mice. (M) A schematic representation of a bacterial translocation axis to vagus nerve and the brain is demonstrated in Paigen diet-fed mice (created with Biorender.com). Data presented from minimum 2 independent experiments (n = 3–5/group). Percentage of mice in each group positive for bacterial CFU in the brain (% pos) is indicated in (E) and (L). Statistical significance was determined by Mann–Whitney U (MWU) test (solid lines) and Fisher Exact (FE) test (dashed lines). *P < 0.05, **P < 0.01, ***P < 0.001. Line reflects the geometric mean. The raw data for this Figure can be found in S1, S2, and S3 Data files.

https://doi.org/10.1371/journal.pbio.3003652.g001

As the vagus nerve connects the gut and the brain, we tested the vagus nerve itself for the presence of bacteria. We isolated small numbers of bacteria, including S. xylosus, from the cervical branches of the vagus nerve, but not the spinal cord (Figs 1K and S2G–S2I), indicating that the localization of these bacteria to the vagus nerve was specific. Similar to brain samples, both quantitative and qualitative analyses were used to determine statistical significance and a positive association between the vagus samples harboring bacteria from the Paigen diet- versus Control diet-fed mice (Fig 1K). Low numbers of S. xylosus (though not reaching statistical significance) were also detected in the nodose ganglia, the inferior ganglia of the vagus nerve, in Paigen diet-fed mice (S2H Fig). To test whether the vagus nerve could serve as a route of bacterial translocation to the brain, we next disrupted the right cervical vagus nerve by vagotomy (visual confirmation of vagotomy provided in S5 Fig) [30,31]. Vagotomized and sham-operated mice were subsequently fed Paigen diet. We observed that vagotomized mice fed Paigen diet harbored ~20-fold lower levels of bacteria in the brain than sham controls (further, 5 out of 15 vagotomized mice had no detectable bacteria in the brain) (Fig 1L). Importantly, vagotomy did not alter intestinal barrier permeability or blood–brain barrier permeability, nor the overall bacterial load in the feces or ileum (S5C–S5F Fig). Moreover, gut microbiome composition in vagotomized mice fed Paigen diet closely matched that of sham-operated controls, as demonstrated by principal component analysis (PCA) of 16S rRNA gene sequencing (S5G Fig). Since vagotomy was performed only on the right cervical vagus nerve (bilateral cervical vagotomy would be fatal), the left cervical vagus nerve remained intact and therefore could have served as a route by which bacteria were still able to translocate to the brain (S5 Fig). Indeed, bacteria were cultured from both the right and left cervical vagus nerves in Mdr2^−/− mice fed Paigen diet (S5A Fig). While we cannot rule out the contribution of other routes, these data indicate that bacterial translocation from the gut to the brain can occur via the vagus nerve in the absence of any observable blood–brain barrier permeability (Fig 1J and 1M).

To further investigate whether bacterial translocation to the brain might occur from the gut, we tested whether bacteria in the brain closely matched those isolated from the gut by whole genome sequencing of bacterial isolates from fecal pellets, the ileum, and the brain of Paigen diet-fed Mdr2^−/− mice. The results showed that S. xylosus isolated from the fecal samples, ileum, and the brain of the same mouse exhibited 99.9982% average nucleotide identity (ANI) (S2 Table). Similarly, S. sciuri isolates from these samples exhibited 99.9977% ANI and E. faecalis isolates exhibited 99.9989% ANI. These data further demonstrate that the bacteria isolated from the brain match those from the gut. Taken together, these results indicate that dietary modulation can cause alterations to gut microbiome composition and gut barrier permeability, leading to translocation of staphylococci and enterococci to the brain in Mdr2^−/− mice.

Intestinal microbiome composition dictates which bacteria translocate to the brain

To further explore the link between intestinal bacteria and those that translocate to the brain, we perturbed the microbiome in Mdr2^−/− mice fed Paigen diet by administration of a cocktail of antibiotics (vancomycin, 500 mg/L; ampicillin 1 g/L; neomycin, 1 g/L; and metronidazole, 1 g/L) in the drinking water. While Staphylococcus species were highly enriched in the ileum and fecal pellets of Paigen diet-fed mice, the addition of antibiotic treatment (Abx) instead led to the enrichment of Paenibacillus cineris (P. cineris) (S6 Fig). Correspondingly, Paigen diet-fed, antibiotic-treated mice rarely harbored Staphylococcus sp. in the brain, but rather, P. cineris was the predominant bacterium detected in the brains of these mice (4 out of 10 were positive for bacteria). Although P. cineris was significantly enriched in the fecal samples and ileum of antibiotic-treated and Paigen diet-fed Mdr2^−/− mice, due to the modest number of mice with P. cineris in the brain (40%), total CFU counts in the brain were not statistically significantly different between mice fed Paigen diet or Control diet. P. cineris is an environmental bacterium which has been reported in human sputum [32,33]. It is surprising yet reasonable that we detected it after antibiotic treatment, which significantly altered the gut microbiome. Similar to the previous data with S. xylosus in mice fed Paigen diet in the absence of antibiotics, we could not detect P. cineris in the blood, other systemic organs, or the spinal cord of mice fed Paigen diet and treated with antibiotics (S6 Fig). In addition, antibiotic treatment did not further exacerbate the intestinal permeability caused by Paigen diet or increase blood–brain barrier permeability (3.5 ± 1.0 versus 3.0 ± 1.4). These data further implicate the gut as the source of brain-localizing bacteria, demonstrating that a perturbation that changes the composition of the microbiome (i.e., diet and antibiotic use) correspondingly changes the bacteria that localize to the brain in the setting of intestinal barrier permeability.

To determine whether bacterial translocation to the brain occurs in mouse backgrounds other than Mdr2^−/−, we fed wild-type C57BL/6 (B6) mice Paigen diet for 2 weeks. Although total CFUs were moderately decreased in fecal pellets and unchanged in the ileum, Paigen diet led to enrichment of S. xylosus and E. faecalis in the feces and ileum (Figs 2A, 2B, and S7). Similar to Mdr2^−/− mice, we were unable to detect bacteria in the blood, or blood–brain barrier permeability, in B6 mice fed Paigen diet. Antimicrobials such as CRAMP (the murine ortholog of human LL-37) and lysozyme can be induced in the blood when bacteria are in the blood, but we did not observe any such increases in B6 mice fed Paigen diet (S8A and S8B Fig). Although we did not observe bacteria in the blood or most systemic organs, we did detect both S. xylosus (statistically significant) and E. faecalis (statistically non-significant due to the low percentage of mice with this bacterium in the brain) in the brain of Paigen diet-fed B6 mice (Figs 2C, 2D, and S7). Taken together (summarized in Box 2), these data further support our conclusion that bacteria are not present in the blood in mice which do have bacteria in the brain, in the models examined here. Genomic sequencing revealed 99.9989% ANI between S. xylosus isolates, and 99.9979% ANI between E. faecalis isolates, present in the feces or ileum and those that translocated to the brain of Paigen diet-fed B6 mice (S2 Table). To test whether the vagus nerve is a route of bacterial translocation to the brain of Paigen diet-fed B6 mice, similar to Mdr2^−/− mice, we performed right cervical vagotomy. 47% of vagotomized and Paigen diet-fed B6 mice (9 out of 19) harbored bacteria in the brain compared to 75% of sham surgical controls (15 out of 20), and the vagotomized mice that did have bacteria in the brain had on average ~3–4-fold lower levels compared to sham controls (Fig 2E). These data demonstrate that bacterial translocation to the brain via the vagus nerve is not restricted to Mdr2^−/− mice fed Paigen diet, but also occurs in wild-type mice of another genetic background (B6) when fed Paigen diet.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Gut microbiome composition dictates which bacteria translocate to the brain.

(A–D) Total CFU of bacteria (A) per gram of fecal pellet, (B) per gram of ileum, (C) per cervical vagus nerve (~2 cm), and (D) per brain of C57BL/6 (B6) mice fed Paigen diet. (E) Total CFU/brain following unilateral right cervical vagotomy of B6 mice fed Paigen diet. (F–H) Kinetics of CFU of bacteria (F) per cervical vagus nerve (~2 cm), (G) per brain, and (H) intestinal permeability in Paigen diet-fed B6 mice. (I–K) CFU of E. cloacae per gram of (I) fecal pellet, (J) ileum, and (K) per brain. (L) Agarose gel image showing bar-coded nucleotide in E. cloacae from the fecal and brain isolates after PCR amplification. WT E. cloacae were used as negative control (NC), and while E. cloacae (JC96) were used as positive control (PC). Percentage of mice in each group positive for bacterial CFU in the brain (% pos) is indicated (D, E, G, K). Statistical significance was determined by Mann–Whitney U test (MWU; solid lines) and Fisher Exact (FE; dashed lines) test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Line reflects the geometric mean. The raw data for this Figure can be found in S1 and S2 Data files.

https://doi.org/10.1371/journal.pbio.3003652.g002

We next investigated the kinetics of bacterial translocation from the gut to the brain during Paigen diet feeding. Interestingly, in B6 mice fed Paigen diet, we detected bacteria in the vagus nerve of one mouse on day 2 (out of 10) and 4 mice on day 4 (out of 10; although not statistically significant), but none in the brains of those mice at the respective timepoints (Fig 2F and 2G). At day 6, we continued to detect bacteria in the vagus nerve and only then, first detected bacteria in the brain. An increase in gut barrier permeability was observed at days 2, 4, and 6 during Paigen diet feeding (Fig 2H). The levels of bacteria cultured from the vagus nerve and brain on day 6 were similar to those on day 14 and even on day 56 (S10 Fig). The bacteria were not detected in the blood, heart, kidney, spleen, lungs, CSF, meninges, or spinal cord on days 4 or 6 in B6 mice fed Paigen diet (S11 Fig). Although the results were not statistically significant, these temporal data provide further anecdotal support for the vagus nerve being a conduit for bacterial translocation to the brain since bacteria were only detected in the brain after their prior detection in the vagus nerve.

Importantly, no changes in whole gut transit time or fecal moisture content, parameters of gut function, were observed in B6 mice fed Paigen diet (S12A and S12B Fig). However, consistent with the increased gut barrier permeability observed, histopathological analysis revealed the disruption of the ileal mucosal lining in B6 mice fed Paigen diet (S12C Fig). The disruption of the mucosal lining correlated with a ~ 2× reduction in the number of goblet cells (which produce mucus) per crypt/villus in B6 mice fed Paigen diet compared to those fed a Control diet (S12D and S12E Fig). These data provide a rationale for why gut barrier permeability is increased in mice fed Paigen diet, since these mice exhibit disruptions to the mucus layer which normally contributes to blocking bacterial translocation out of the intestine.

To elucidate whether we could exogenously alter which bacteria reach the brain, we first disrupted gut bacterial composition by treating B6 mice with a cocktail of antibiotics. We subsequently gavaged Enterobacter cloacae (strain JC96 with an engineered, non-natural barcoded nucleotide sequence) which is not observed in the gut microbiome of the B6 mice, into the antibiotic-treated mice and fed them Paigen diet. E. cloacae were detected in fecal samples and the ileum on day 5 after gavage (Fig 2I and 2J). On day 8 after gavage, E. cloacae were cultured from the brains of 40% of gavaged and Paigen diet-fed B6 mice (Fig 2K). To further establish the gut as the origin of bacterial translocation to the brain, we confirmed the presence of the specific JC96 DNA barcode in the E. cloacae isolates from both the fecal pellets and brain samples (Fig 2L).

We subsequently monocolonized germ-free (GF) B6 mice by gavaging E. cloacae JC96 with or without Paigen diet feeding. Increased gut barrier permeability was only observed in monocolonized GF mice that were also fed Paigen diet, but not in monocolonized GF mice fed Control diet (Fig 3A). Correspondingly, E. cloacae was cultured in the fecal pellets and ileum of monocolonized GF mice fed either Paigen diet or Control diet (Fig 3B and 3C), however, it was only cultured from the vagus nerve and brain of monocolonized GF mice also fed Paigen diet (although statistically insignificant in the vagus due to the low proportion, 25%, of mice with bacteria in this nerve) (Fig 3D and 3E). These data rule out contamination of brain samples with intestinal bacteria during sample collection, since monocolonized GF mice fed Control diet harbored high levels of E. cloacae in the gut, yet had no detectable bacteria in the brain. Together, these data highlight that intestinal colonization alone is not sufficient to cause bacterial localization in the brain, but it must be accompanied by gut barrier permeability for bacteria to translocate to the brain. The data further demonstrate that the bacteria isolated from the brain are also present in the gut, consistent with the gut being the source of the bacteria that translocate to the brain.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Translocation of monocolonized bacteria to the brain in germ-free mice, detection of bacterial DNA by ddPCR, and diet reversal.

(A–E) Germ-free (GF) mice were monocolonized (mono) by gavaging E. cloacae with or without Paigen diet feeding, and (A) intestinal permeability and (B-E) bacterial loads (CFU) were determined (B) per gram of fecal pellet, (C) per gram of ileum, (D) per cervical vagus nerve (~2 cm) and (E) per brain. (F) Mdr2^−/− mice were monocolonized with E. cloacae (JC96 strain) and fed Paigen diet. Brains were harvested, and microbial DNA from the brain was isolated using HostZERO Zymo and QIAamp Microbiome Kit. Using primer probe pairs designed for the JC96 barcode, a PCR mix was made using ddPCR supermix for probes (No dUTP) with forward and reverse primers, EcoRI, and DNA template to directly quantitate bacterial DNA. Non-template control (NTC), Zymo kit and Qiagen kit controls, and DNA isolated from Klebsiella oxytoca (K. oxytoca), S. xylosus, E. faecalis were used as non-specific controls. DNA isolated from parenteral Enterobacter with no barcode (JC5) was used as negative control, while DNA isolated from Enterobacter with barcode (JC96) was used as a positive control. For control and experimental groups, DNA isolated from JC96 monocolonized mice fed Control diet and Paigen diet were used. After the PCR, the droplets were read in a QX200 Droplet Digital System by the direct quantification method. The figure was generated using the QuantaSoft Analysis Pro Software. A detailed description of the PCR reaction and primer probes provided in the Materials and methods section. (G–I) Paigen diet was fed to 12-week-old Mdr2^−/− mice for 9 days (“Paigen” group), before reversal to conventional diet for 14 (“D14”) or 28 (“D28”) days. (G) Intestinal permeability in Paigen diet-fed mice and diet reversal mice on D14 and D28. (H, I) CFU of S. xylosus (H) per gram of ileum and (I) per brain. Data presented from minimum 2 independent experiments (n = 2–3/group). Percentage of mice in each group positive for bacterial CFU in the brain (% pos) is indicated in (E) and (I). Statistical significance was determined by Mann–Whitney U test (MWU; solid lines) and Fisher Exact (FE; dashed lines) test or one-way analysis of variance, where applicable. *P < 0.05, **P < 0.01, ***P < 0.001. Line reflects the geometric mean. The raw data for this Figure can be found in S1 and S2 Data files.

https://doi.org/10.1371/journal.pbio.3003652.g003

Detection of bacterial DNA in the brain by PCR

We next attempted to corroborate our bacterial culture data with molecular detection of bacterial DNA in the brain, following the most recent Consensus Statement on guidelines for low-biomass microbiome studies to ensure quality, robustness, and reproducibility (multiple primer sets, different kits for isolation of DNA, and many other controls summarized in Box 3) [34]. We first colonized the gut of B6 mice with Enterobacter strain JC96 and fed them Paigen diet. Brains were harvested and microbial DNA was isolated using two independent kits (HostZERO Microbial DNA Kit, Zymo, or QIAamp DNA Microbiome Kit, Qiagen). PCRs specific to the DNA barcode in Enterobacter JC96 were performed using distinct primer sets (JCP51 with JCP53, or JCP429 with JCP430). While the presence of bacteria in brain samples was confirmed by our culturing method, we were unable to amplify any DNA products by direct PCR assay of the brain samples (S3 Table, S13A Fig). We therefore determined the limit of detection (LOD) for the PCR methodology. On pure cultures of JC96, the LOD was determined to be ~10³ bacterial cells. However, when brain tissue homogenate was spiked with an indicated known number of bacterial CFU, the LOD for the PCR assay was determined to be roughly one log higher, at ~10⁴ bacterial CFU (S4 Table, S13B Fig). Since we only observe low numbers of bacteria in the brain in the models studied here, and this never reaches 10⁴ CFU, we therefore reasoned that it may be possible to facilitate detection of the bacteria by standard PCR, if the bacteria in the brain homogenates were first cultured and expanded in Mueller Hinton broth (MHB) for a minimum of 3 hours at 37 °C. When this approach was implemented, the sensitivity of the LOD was improved by ~100-fold, facilitating the detection of just 10 initial bacteria (prior to culture) by conventional PCR on brain homogenates from Paigen diet-fed mice after culturing and expansion in MHB. Detection of bacteria in the brain by PCR after MHB enrichment from Mdr2^−/− and B6 mice colonized with JC96 and fed Paigen diet is shown (S3 Table, S13C and S13D Fig). Moreover, we used a second set of primers for the detection of E. cloacae strain JC96 in the brain of B6 mice using MHB enrichment by standard PCR (S14 Fig).

Consistent with the qPCR data which indicating difficulty in detecting low numbers of bacteria, we have not thus far been able to detect bacteria in the brain or vagus nerve by imaging. Therefore, to attempt to detect bacteria in the brain without the need for in vitro expansion, we turned to droplet digital PCR (ddPCR), one of the most sensitive PCR technologies available [35–37]. We employed this approach on brains from Mdr2^−/− mice (because we observed the highest level of bacterial load in the brain in these mice) gavaged with Enterobacter strain JC96 following Abx-treatment and fed Paigen diet. Microbial DNA from the brain was isolated using two different DNA isolation kits. As shown in Fig 3F, ddPCR specifically detected the DNA of barcoded Enterobacter JC96. Detection of this barcode in the brain indicated that the detected DNA came specifically from the bacterial strain with which mice were colonized and could not have come from other naturally occurring bacteria in the mice. Importantly, the primers that detected the DNA barcode specific to JC96 did not amplify a PCR product from DNA of the parental JC5 strain which is devoid of the barcode, and similarly did not amplify a product from unrelated negative control strains of Klebsiella oxytoca, Staphylococcus xylosus, or Enterococcus faecalis (Fig 3F). In addition, brains from Mdr2^−/− mice colonized with JC96 but fed a regular diet, which did not lead to bacterial translocation to the brain, were negative by ddPCR (Fig 3F). These data demonstrate the presence of bacterial DNA in the brain of mice which harbor bacteria as assayed by culturing, corroborating the culture data with molecular detection.

Restoration of conventional diet reverses bacterial localization in the brain

Because mice fed a Paigen diet developed increased intestinal barrier permeability and bacteria translocated to their brains, we tested whether these phenotypes could be reversed upon restoration of a conventional diet. We first fed Paigen diet to Mdr2^−/− mice for 9 days (“Paigen” group), before subsequently reversing their diet back to the Control diet, and then testing their phenotypes 14 days (D14) and 28 days (D28) later. Restoration to a conventional diet decreased intestinal barrier permeability by 4-fold on D14 and D28 (Fig 3G). In addition, gut microbiome composition on D28 was reversed and closely matched that of Control diet-fed animals (S15 Fig). With only two exceptions, levels of S. xylosus were below the detectable limit in both the ileum and brain on D14 (Fig 3H and 3I). Similar to these results from Mdr2^−/− mice indicating that bacterial localization to the brain can be a reversible process, when B6 mice were fed Paigen diet for two weeks (“Paigen” group) and the diet was subsequently reversed to the Control diet, there was a significant decrease in intestinal barrier permeability on day 7 (D7) and the bacterial loads in the brain on day 21 (D21) and day 28 (D28) (S16A and S16B Fig). These data demonstrate that the localization of bacteria in the brain can be transient as diet reversal led to a reduction of detectable bacteria in the brain.

Bacterial localization in the brain of mouse models of neurologic conditions

Recent evidence implicates the disruption of the gut microbiome and an increase in gut barrier permeability as being associated with neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease, and neurodevelopmental conditions such as autism spectrum disorder [1–4,38–40]. We employed mouse models for each condition (Alzheimer’s—APP/PS1 WSB.Cg-Tg (APPswe,PSEN1dE9) 85Dbo/How; Parkinson’s—LRRK2 knock-in (B6.Cg-Lrrk2^tm1.1Hlme/J; and autism spectrum disorder—BTBR T+ Itpr3^tf/J BTBR) to determine if changes in gut microbiome composition and gut barrier permeability might be linked to bacterial translocation to the brain in these mice. APP/PS1 mice exhibited enriched Ligilactobacillus, Streptococcus, Clostridium, Lachnospiraceae, Roseburia, and Tyzzerella in comparison to wild-type B6 mice (Figs 4A and S17), while LRRK2 mice exhibited marked enrichment of Lactobacillus, Eubacterium, Staphylococcus, Enterorhabdus, Coriobacteriia, and Flavonifractor (Figs 4A and S18). BTBR mice showed a broader enrichment of Firmicutes, Lachnospiraceae, Corynebacterium, Anaeroplasma, Monoglobus, Fusimonas, Murimonas, Tyzzerella, Erysipelotrichaceae, Enterobacter, and Robinsoniella in comparison to B6 mice (Figs 4A and S19). We observed gut barrier permeability to be a common phenotype in these mouse models of Alzheimer’s, Parkinson’s, and autism spectrum disorder (Fig 4B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Bacterial translocation to the brains of mouse models of neurologic conditions.

(A) Fecal pellets collected from WT, APP/PS1, LRRK2, and BTBR mice from 8- to 15-week-old mice were processed for 16S rRNA gene sequencing. Cladogram representation showing the phylogenetic relationship between bacterial families among WT, APP/PS1, LRRK2, and BTBR mice. (B) Intestinal permeability in WT, APP/PS1, LRRK2, and BTBR mice. (C–E) Total CFU of bacteria (C) per brain, (D) in the vagus nerve (~2 cm), and (E) per mL of blood. Data presented from minimum 2 independent experiments (n = 2–3/group). Percentage of mice in each group positive for bacterial CFU in the brain or vagus nerve (% pos) is indicated (C, D). Statistical significance was determined by Mann–Whitney U test (MWU; solid lines) and Fisher Exact test (FE; dashed lines) or one-way analysis of variance, where applicable. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Line reflects the geometric mean. The raw data for this Figure can be found in S1, S2, and S3 Data files.

https://doi.org/10.1371/journal.pbio.3003652.g004

Without any dietary intervention (Control diet), small numbers of bacteria were detected in the brain as well as the vagus nerve in these murine models (Fig 4C and 4D). Importantly, bacteria that were isolated from the brain and the vagus nerve in these models were always also detected in the fecal samples and/or ileum of each respective mouse (S5 and S6 Tables). Furthermore, bacteria were not detected in systemic tissues or the blood (S5 Table, Fig 4E) among these mice, and blood–brain permeability indexes were not elevated compared to WT controls (3.4 ± 1.7 for B6, 3.9 ± 0.2 for APP/PS1, 3.3 ± 2.3 for LRRK2, and 3.6 ± 1.8 for BTBR). These findings provide evidence that alteration of gut microbiome composition and increased gut barrier permeability may lead to bacterial translocation to the brain in at least some mouse models of neurodegenerative and neurodevelopmental disease.

Ethics statement

The use of animal models and procedures in this study were approved by the Institutional Animal Care and Use Committee (IACUC) of Emory University (Protocol # PROTO-201700372). To minimize co-housing/cage effects and coprophagy on microbiota composition, we minimized housing density (keeping 2–3 mice per cage) and utilized littermate controls. Mouse strains were kept under similar housing conditions and were fed the same rodent chow, unless stated for Paigen Diet. We found no significant differences in the percentage of mice positive for bacteria in the brain in the same group housed in different cages and also in the same group from different experiments. Based on observations and data collected from multiple independent repeats, we consistently detected the same type of bacteria in the brain under the same particular treatment conditions, with similar frequencies of detection, even when different batches of mice were used. These results support the conclusion that cage effects are unlikely to have influenced the findings.

Mouse strains

C57BL/6J (000664), FVB/NJ (001800) (wild-type controls) and Mdr2^−/+ (FVB.129P2-Abcb4^tm1Bor/J) mice were obtained from the Jackson Laboratory (ME, USA) and housed in a specific pathogen-free environment, per Emory University, NIH, and IACUC guidelines. Mdr2^−/− mice, Mdr2^−/+ mice, and control mice were generated by backcrossing littermates until desired true breeding lines were established and confirmed by PCR. BTBR T+ Itpr3^tf/J (002282), APP/PS1 WSB.Cg-Tg(APPswe,PSEN1dE9)85Dbo/How (038560), and LRRK2 knockin (B6.Cg-Lrrk2^tm1.1Hlme/J (030961) mice were purchased from the Jackson Laboratory and housed in a specific pathogen-free environment per Emory University, NIH, and IACUC guidelines. A standard chow (control; Labdiet 5001) or Paigen’s Atherogenic Rodent diet (Research Diets D12336 consisting of fat 35 kcal%, protein 20 kcal%, and carbohydrate 45 kcal% with cholesterol 1.25%, and sodium cholic acid 0.5%. Fat was from soybean oil 5%, cocoa butter 7.5%, coconut oil 3.5%), was fed to Mdr2^−/− mice (both male and female) (9 days) or B6 mice (14 days). In some experiments, mice were given a cocktail of antibiotics (neomycin sulfate, 1 g/L; ampicillin 1 g/L; vancomycin 500 g/L; and metronidazole 1 g/L) in drinking water, while on Paigen diet.

Intestinal permeability assay

Mice were fasted for 8–10 hours and orally gavaged with 100 μL of 160 mg/mL FITC-dextran (4 kDa) dissolved in PBS. Blood was collected from mice 4 hours post-gavage by submandibular bleed, and serum was collected after centrifugation. Dilutions of FITC dextran solution ranging from 15.6 ng/mL to 8,000 ng/mL were made for the standard curve. The fluorescence of diluted serum and dilutions for the standard curve was measured in a fluorimeter after excitation of the samples at 488 nm and emission at 521 nm. The amount of FITC-dextran in serum was determined by the standard curve.

Whole gut transit time

To determine whole gut transit time, mice were fasted overnight, followed by removal of water for 1 hour in the morning. The mice were single-housed and gavaged with 100 μL of 5% carmine red dissolved in 1.5% methyl cellulose. The time between gavage and the time at which the first red fecal pellet was calculated as the whole gut transit time.

Fecal moisture content

For determination of fecal moisture content, fecal pellets were collected, and their weights were noted. This was followed by drying the feces for 24 hours at 95 °C. The difference in the weight of feces before and after drying was considered the moisture content in the feces.

Blood–brain barrier permeability index

The blood–brain barrier permeability index (BPI) was calculated using the FITC tracer as described previously [64]. Briefly, the mice were injected intraperitoneally with 100 μl of 2mM FITC dextran (3KDa) solution in PBS. After 20 min, mice were euthanized, and blood was collected by cardiac puncture, followed by intracardiac perfusion with 20–40 mL of PBS. The brain was collected in 1 mL PBS and homogenized. The homogenate was centrifuged at 15,000g for 20 min, and the fluorescence intensity (F.I.) of FITC in the supernatant, along with F.I. of FITC in serum, was measured using a fluorescence plate reader. The F.I. per gram of the brain was determined by dividing the F.I. of the brain homogenate by the brain weight (normalized brain F.I.). Similarly, F.I. per mL of serum was calculated by dividing the F.I. of the serum by the serum volume used for fluorescence calculation (normalized serum F.I.). The BPI was calculated by dividing the normalized brain F.I. by normalized serum F.I. C57BL/6 mice injected intraperitoneally with 3 mg/kg Salmonella enterica LPS was used as a positive control [65]. LPS-injected C57BL/6 mice had a BPI of 16.6 ± 3.6 at 24 hours.

Blood culture

Mice were euthanized by CO₂ asphyxiation. Blood was collected terminally from the mice by cardiac puncture and put in EDTA/Heparin Blood collection tubes. 100 μL of the blood was put in 1 mL of BacT/ALERT FA plus aerobic media and incubated for up to 7 days at 37 °C with 5% CO₂ with subculture on blood agar on day 2 and day 7 to culture any bacteria present in the blood. Separately, we assessed the limit of detection (LOD) of the BacT/ALERT FA plus aerobic blood culture media. Staphylococcus xylosus and Enterobacter cloacae strains were cultured in vitro, diluted, inoculated into the blood culture media and incubated as above. By measuring the input CFU, we found that these culture conditions can detect as low as 1 CFU of bacteria.

Brain and other organ culture

Mice were euthanized by CO₂ asphyxiation. The brain was dissected first from all mice and collected under aseptic conditions. Following the brain, heart, lungs, kidneys, spleen, and ileum tissues were collected under aseptic conditions. Different sets of tools were used for different organs and each mouse. After each use, the instruments were sterilized by dipping in 70% ethanol and washing them in sterile PBS. The organs were collected in either plastic-capped sterile culture tubes with 1 ml of sterile 1× PBS or in 2 mL omni bead beater tubes with 1.5 mm ceramic beads. The culture tubes were weighed before and after organ collection to calculate the weights of the organs. The organs were homogenized, and the probe of the homogenizer was washed 4 times (2 washes with sterile water, 1 wash with 70% ethanol, and 1 wash with sterile PBS) in between the samples for the plastic-capped culture tubes. Bead-beater tubes were shaken in a Bead Ruptor4 for 20 seconds with speed 5. Depending on the tissues, 150–500 μL of the homogenate was plated on a blood agar plate and incubated at 37 °C with 5% CO₂. The total CFU of bacteria in per gram of tissues (CFU/g) was calculated by counting the number of colonies on the blood agar plate and factoring in the corresponding volume/dilution of homogenate and weight of tissues. We picked one morphologically similar colony of each type, from each sample (e.g., brain, ileum, feces) in each experiment. We identified the colonies by MALDI-TOF mass spectrometry at the Clinical Microbiology Laboratory of Emory University Hospital. The individual species count data are therefore estimates since not every colony on every plate was assessed using this approach.

Vagus nerve and spinal cord culture

Mice were euthanized by CO₂ asphyxiation. The left and right cervical vagus nerves running along the neck (~2 cm in length) were dissected under a dissection microscope aseptically and placed in sterile screw cap tubes with 1.5 mm ceramic beads and 500 μL of PBS. Similarly, the spinal cord was isolated from the trunk aseptically and put in a screw cap tube. The samples were homogenized in a bead homogenizer at a speed of 5 m/s for 20 seconds twice. 200 μL of the homogenate was plated on a blood agar plate and incubated at 37°C with 5% CO₂. The total CFU of bacteria per vagus nerve (~2 cm) or in gram of spinal cord (CFU/g) was calculated by counting the number of colonies on the blood agar plate and factoring in the corresponding volume/dilution of homogenate. Bacterial colonies (one of each morphological type from each sample) isolated from tissue homogenates were identified by MALDI-TOF mass spectrometry at the Clinical Microbiology Laboratory of Emory University Hospital. The individual species count data are therefore estimates since not every colony on every plate was assessed using this approach.

Cervical vagotomy

Mice were anesthetized using 4% isoflurane, and the surgical plane was maintained with 1.5% isoflurane. 1 mg/kg of buprenorphine E.R. was injected subcutaneously as a presurgical analgesic. Mice were placed in the supine position, and the neck area was shaved and cleaned with alcohol and betadine solution. A 2 cm incision was made in the skin in the middle of the neck using a scalpel. Under a dissecting microscope, the connective tissue layer was separated using a pair of forceps to expose mandibular glands. The connective tissue layer was further separated to expose the sternohyoid, sternomastoid, and omohyoid muscles. The sternomastoid muscle of the right side was retracted to visualize the pulsating carotid artery and the right vagus nerve. The right vagus nerve was carefully separated from the carotid sheath, and a portion of the vagus nerve was cut with the help of micro-dissecting scissors. The skin was sutured, and mice were allowed to recover. SHAM surgery was performed, where the right vagus nerve was separated from the carotid sheath but not cut. Of note, bilateral cervical vagotomy was not possible since this would be fatal.

Enterobacter cloacae gavage and detection

Enterobacter cloacae isolate was described previously [66] and was genetically modified for this study by integrating a ~ 5 kb sequence (bar-coded) in its chromosome (to generate strain JC96). E. cloacae from an overnight culture washed in sterile PBS (5 × 10⁷ cfu) was orally gavaged into C57BL/6 mice previously treated with a cocktail of antibiotics. Following gavage, mice were subsequently fed Paigen diet. Colonization of E. cloacae was detected in the gut on day 5 after gavage by culturing fecal pellet and the ileum tissue using a standard bacterial culture method. The total CFU of bacteria in per gram of feces or tissues were calculated by counting the number of colonies on the blood agar plate and factoring in the corresponding volume/dilution of homogenate and weight of tissues. Bacterial colonies isolated from feces, ileum, and brain tissue homogenates were identified by MALDI-TOF mass spectrometry at the Clinical Microbiology Laboratory of Emory University Hospital. The presence of bar-coded sequence was confirmed by PCR amplification using oligonucleotides: JCP429: 5′-AGTGGAGAGGGTGAAGGTGA-3′ and JCP430: 5′-CAATGTTGTGGCGAATTTTG-3′. These oligonucleotides target an internal region of the bar-coded sequence amplifying 437 bp. Wild-type E. cloacae was used for negative control (NC), whereas the genetically modified E. cloacae (JC96) was used as positive control (PC).

Monocolonization of mice with E. cloacae by oral feeding

B6 mice were treated with a cocktail of antibiotics (Ampicillin, 500 mg/L; Neomycin, 500 mg/L; Metronidazole, 500 mg/L; and Vancomycin, 250 mg/L) for 3 days, and the depletion of microbiota was confirmed by using standard bacterial culture of feces. The mice were fasted for 4 hours prior to feeding a bacterial suspension of E. cloacae JC96 containing 10⁹ CFU/mL in 2% sucrose solution ad libitum for 15 min. The mice were returned to their cage and continued with ampicillin (500 mg/L in drinking water) treatment to maintain JC96 colonization, as JC96 is ampicillin-resistant. Colonization was confirmed at day 4 after JC96 feeding by bacterial culture of feces. After confirmation of colonization, mice were fed the Paigen diet and follow-up experiments were performed as described above.

Isolation of microbial DNA from the brain

Two different kits (QIAamp DNA Microbiome Kit by Qiagen and HostZERO Microbial DNA Kit from Zymo Research), specialized to isolate microbial DNA from low-biomass samples, were used to isolate microbial DNA from the mouse brain. Both kits use two steps. In the first step, host DNA depletion is performed to remove eukaryotic host DNA by limited permeabilization of the host cells. This is followed by the lysis of prokaryotic cells and the isolation of the prokaryotic DNA. The microbial DNA was isolated following the manufacturer’s protocols.

PCR from microbial DNA isolated from the brain

Microbial DNA isolated using the microbiome kits mentioned above was used for PCR. PCR was done for the detection of JC96 DNA using two sets of primers, one targeting the barcode (JCP429, JCP430) and the other targeting a region covering the barcode and Enterobacter DNA (JCP51, JCP53) using the GoTaq MasterMix. The following primer sequences were used to detect JC96 from the microbial DNA isolated from the brain.

1. JCP51: 5′ GCCAGTTTCAGCAAGGTTTC 3′
2. JCP53: 5′ GCTCTTCCGCTTCCTCGCTC 3′
3. JCP429: 5′ AGTGGAGAGGGTGAAGGTGA 3′
4. JCP430: 5′ CAATGTTGTGGCGAATTTTG 3′

Initial denaturation was performed at 95 °C for 5′ followed by 30 cycles of denaturation at 95 °C for 30″, annealing at 52°C (for JCP51, JCP53) or 55 °C (for JCP429, JCP430) for 60″, and extension at 72 °C for 90″ (for JCP51, JCP53) or 30″ (for JCP429, JCP430), followed by final hold at 72 °C for 5′.

For the samples where microbial DNA could not be detected either due to the absence of JC96 DNA or due to lower JC96 DNA levels than the LOD, we used an enrichment strategy. The brain homogenate for those samples was incubated with added MHB overnight at 37 °C, followed by microbial DNA isolation described in the previous step. Uninoculated MHB was included as a negative control.

Droplet digital PCR (ddPCR) for detection of bacterial DNA

Next, we used ddPCR, a more sensitive and high-throughput PCR technique [35–37], to quantify bacterial DNA in the brain without ex vivo enrichment of the bacteria. Briefly, primer probe pairs were designed for the JC96 barcode and a PCR mix was made using ddPCR supermix for probes (No dUTP) following BioRad guideline with forward and reverse primers, probes, EcoRI, and template. A droplet generator was used to divide the PCR mix into droplets using Droplet Generation Oil for Probes from Bio-Rad. The PCR was performed at 95 °C for 10′ followed by 40 cycles of denaturation at 94 °C for 30″, annealing at 60 °C for 60″, and extension at 72 °C for 30″, followed by a final hold at 98 °C for 10′. After the PCR, the droplets were read in a QX200 Droplet Digital System by the direct quantification method. The figure was generated using the QuantaSoft Analysis Pro Software.

1. Forward Primer: 5′ GAGGGTGAAGGTGATGCTAC 3′
2. Reverse Primer: 5′ CGGGAAAAGCATTGAACACC 3′
3. Probe: 5′ HEX-CCTGTTCCA ZEN TGGCCAACACTTGTCAC 3′ IB FQ

Germ-free mouse experiments

The experiments were conducted at the Emory Gnotobiotic Animal Core. Five week old germ-free B6 mice were monocolonized with Enterobacter cloacae JC96. Intestinal colonization was confirmed through fecal samples by culturing and detecting total CFUs in the feces by a standard bacteriological culture as described above, followed by PCR using JCP429 and JCP430 primers and MALDI-TOF identification of the cultured bacteria. After colonization, the mice were fed Paigen diet for two weeks. Mice were harvested aseptically, and gut barrier permeability was assessed along with bacteriological culturing of various samples such as brain, vagus nerve, ileum, heart, spleen, lungs, kidney, meninges, spinal cord, and CSF. Any bacteria detected in the brain were further identified by PCR using JCP429 and JCP430 primers as described above to confirm the presence of the barcode and MALDI-TOF to confirm the species identification.

CRAMP and lysozyme ELISA

The serum was diluted 1:2 for CRAMP (the murine ortholog of human LL37) ELISA (LS Bio, Cat. # LS-F32286) and 1:500 for lysozyme ELISA (Novus Biologicals, Cat. # NBP2-68058),added to the precoated wells and incubated for 90 min at 37 °C, followed by incubation with biotinylated detection antibody for 60 min at 37 °C. After washing the plate, an HRP-streptavidin conjugate was added to the wells and incubated for 30 min at 37 °C. The plates were washed, and the TMB substrate was added. Once blue color was developed, the reaction was stopped by adding the stop solution in the kit, and OD450 was measured in an Agilent BioTek Synergy H1 microplate reader. The concentration of LL37 and lysozyme was calculated in the serum by a standard curve of LL37 and lysozyme proteins provided in the kit.

Alcian Blue Periodic Acidic-Schiff (AB-PAS) staining of the gut

B6 mice were fed either regular or Paigen diet for 14 days and harvested. The small intestine was removed, and the intestinal contents were gently squeezed out of the gut using forceps. The gut tissues were then fixed in 10% formalin solution for 24 hours, followed by preserving them in 70% ethanol. The gut tissues were then dehydrated and embedded in paraffin. 5 μm thin sections were cut in a microtome, and the sections were put on a charged slide. For staining, the slides were deparaffinized, rehydrated, and then stained with alcian blue and Schiff’s reagent to stain acidic and neutral mucins, respectively. The slides were then counterstained with hematoxylin. The slides were then dehydrated and mounted using a xylene-containing mounting medium. The images were acquired in a brightfield microscope.

16S rRNA gene sequencing

Fecal samples were collected from Mdr2^−/− mice on Control diet or Paigen diet for DNA isolation using Qiagen Powersoil Pro DNA extraction method at Emory Integrated Genomic Core (EIGC) of Emory University School of Medicine. Libraries were prepared using the 16S rRNA gene sequencing library preparation (Illumina workflow). In this workflow, 16S rRNA gene sequencing amplicons were generated using KAPA HiFi HotStart ReadyMix and primers specific to the 16S rRNA gene sequencing (V3-V4) region (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGG NGGCWGCAG-3′; 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′). Purified amplicons were indexed with Illumina Nextera XT Index Primers and purified with Ampure XP beads. 16S rRNA gene sequencing libraries were pooled in equal amounts based on fluorescence quantification. Final library pools were quantitated by quantitative PCR and sequenced on an Illumina MiSeq using MiSeq v3 600 cycle chemistry. For data analysis, paired-end reads were assigned to samples based on their unique barcodes. After trimming of primer sequences by cutadapt (v1.9), sequencing reads were merged into tags based on overlapping regions using FLASH (v1.2.8). Quality filtering was performed by fqtrim (v0.94) with sliding windows to retain sequences having average quality scores ≥20 and longer than 100 bp in size. Chimeric sequences were discarded using Vsearch (v2.3.4). After dereplication using DADA2 [67], ASV feature sequences were obtained and quantified. Singleton ASVs were excluded from further analysis. Species annotations of ASVs were performed by SILVA (release 132, https://www.arb-silva.de/documentation/release138/) [68] and NT-16S. Beta diversity was calculated based on relative abundance of ASVs [69]. QIIME2 was used for beta diversity analyses, including Chao1, Observed_species, Goods_coverage, Shannon, and Simpson indices. Graphs were generated by ggplot2 (R package v3.5.2). Differential analysis was performed based on relative abundance values. Linear Discriminant Analysis Effect Size (LEfSe) predictions for bacterial groups showing linear marked enrichment (LDA score) and cladogram representation were generated by using R package v3.5.2 [70]. MWU was used for comparisons between 2 groups. Kruskal–Wallis test was used for comparisons among 3 or more groups. p values were corrected for the false discovery rate following Benjamini–Hochberg procedure. A species with p < 0.05 was considered significantly differential species. Bioinformatics services and analyses including diversity, taxonomy, and differential analyses for bacterial families were provided by LC Sciences (Houston, Texas).

Genomic sequencing of bacterial isolates

Bacterial isolates were streaked on blood agar plates and processed for DNA extraction, sequencing, and variant calling at SeqCenter (https://www.seqcenter.com). Illumina sequencing (200 Mbp; 1.33M reads) or small nanopore combo (minimum 300Mbp long reads, 400Mbp Illumina reads) was performed. One of the three isolates from the same mouse (brain, ileum, or feces) was assembled and compared to the Illumina-generated 2*151 bp paired-end read data from the remaining isolates from the same mouse for variant calling using BreSeq (version 0.38.1) under default settings [71]. The number of predicted mutations (single-nucleotide polymorphism) were counted and ANIs were calculated as described previously [72].

Statistical analysis

Data were analyzed using MWU test or FE test or one-way analysis of variance (ANOVA), where applicable, using Graph Pad Prism software (GraphPad Prism, 10.0.2 (171) version). CFUs from brain or vagus tissue homogenates were analyzed for significance using the MWU and FE. The statistical significance for differential analysis based on relative abundance were determined as described in 16S rRNA gene sequencing.