You are what you eat...at least, your brain is

INTRODUCTION

Our gut microbiota is important for many biological functions in the body, including intestinal development, barrier integrity and function (1, 2), metabolism (3, 4), the immune system (5), and the central nervous system (CNS). The effects of the gut microbiota on brain physiology include synaptogenesis, regulation of neurotransmitters and neurotrophic factors such as brain-derived neurotrophic factor and nerve growth factor-A1 (6). However, the development of the CNS also includes the formation of an intact blood-brain barrier (BBB) that ensures an optimal microenvironment for neuronal growth and specification (7). An intact and tightly regulated BBB is also required to protect against colonizing microbiota in neonates during the critical period of brain development (8, 9). It also protects against exposure to “new” molecules and bacterial metabolites due to the postnatal metabolic switch from predominant dependence on carbohydrates during fetal life to a greater dependence on fatty acid catabolism after birth.

The BBB begins to develop during the early period of intrauterine life (10, 11) and is formed by capillary endothelial cells sealed by tight junctions, astrocytes, and pericytes. Tight junction proteins restricting paracellular diffusion of water-soluble substances from blood to the brain (12) consist mainly of transmembrane proteins such as claudins, tricellulin, and occludin, which are connected to the actin cytoskeleton by the zona occludens (ZO-1) (13). Tight junction proteins are dynamic structures and are subject to changes in expression, subcellular location, posttranslational modification, and protein-protein interactions under both physiological and pathophysiological conditions (12). Disruption of tight junctions due to disease or drugs can lead to impaired BBB function, compromising the CNS. Therefore, understanding how BBB tight junction proteins are affected by various factors is important for elucidating how to prevent and treat neurological diseases.

Here, we report that the intestinal microbiota affects BBB permeability in both the fetal and adult mouse brain. Using as a model system germ-free mice that have never encountered a live bacterium and pathogen-free mice that were reared in an environment free of monitored mouse pathogens, we demonstrated that lack of gut microbiota is associated with increased BBB permeability and altered expression of tight junction proteins. Fecal transfer from mice with pathogen-free gut flora into germ-free mice or treatment of germ-free mice with bacteria that produce short chain fatty acids (SCFA) decreased the permeability of the BBB.

RESULTS

The maternal gut microbiota can influence prenatal development of the BBB

First, we characterized BBB permeability of mouse embryos with pathogen-free mothers by administering infrared-labeled immunoglobulin G2b (IgG2b) antibody to dams during timed pregnancies to see whether the antibody was excluded by the BBB or was able to cross the BBB into the brain parenchyma. The qualitative analysis of mouse embryos with pathogen-free mothers showed a shift from a diffuse infrared-labeled antibody signal present within the embryonic brain at E13.5 and E14.5 to a signal confined to the developing vasculature starting at E15.5 to E17.5 (Fig. 1A). This signal was most pronounced in adult offspring of pathogen-free dams (Fig. 1A). The quantitative analysis of the penetration into the fetal brain of infrared-labeled IgG2b antibody injected intravenously into pathogen-free dams supported the qualitative data, showing a decrease at E15.5 to E17.5 (Fig. 1B). In contrast, the analysis of E16.5 brains from fetal mice of germ-free dams showed a diffuse signal from the infrared-labeled IgG2b antibody (Fig. 1C) present in the brain parenchyma (Fig. 1, D and E). Higher-magnification images of the brain showed that the IgG2b antibody was limited only to the vessels in E16.5 fetal mice of pathogen-free dams in contrast to age-matched fetal mice of germ-free dams (Fig. 1, D and E). Because BBB integrity is controlled in part by sealing of the endothelial cells via tight junctions, we determined expression of the main tight junction proteins in brain lysates from E18.5 fetal mice of pathogen-free versus germ-free dams. Expression of the brain endothelial tight junction proteins claudin-5 and ZO-1 was similar between the two groups, whereas the expression of occludin was significantly lower in the brain lysates from E18.5 fetal mice of germ-free dams compared to that in age-matched fetal mice of pathogen-free dams (P < 0.05) (Fig. 1, F and G).

Main Question

Is the BBB more permeable (leaky) during the embryonic development of germ-free mice, compared with pathogen-free mice?

Experimental setup

The authors injected an infrared-labeled antibody into pregnant germ-free or pathogen-free mice (dams) at various timepoints during their pregnancies, then allowed 1 hour for it to reach and spread through the embryonic mice.

The infrared label allowed them to image the location of the antibody within the mouse embryos (Parts A-D).

They sliced the brains of embryonic mice at 16.5 days in utero, then stained the slices with antibodies to label blood vessels, in order to precisely examine the location of the antibody within the embryonic brains (Part E).

Finally, they tested the expression levels of three proteins that form tight junctions, structures that are key for the function of the BBB (Parts F-G). For this, they used Western blotting, a common technique wherein samples of brain are ground up and labeled with antibodies that identify specific proteins, allowing the very sensitive detection of changes of protein expression level.`

BBB Formation

Part A illustrates the development of the BBB. When pathogen-free pregnant dams were injected with an antibody at day 13.5 of their pregnancies, the antibody spread through the entire embryonic brain (see the extensive spreading of the green antibody label in Part A, leftmost image, which shows the head and upper body of an embryonic mouse). This indicates that the BBB is not sufficiently developed to filter out the antibody at this early stage.

In contrast, by day 16.5 and 17.5 of the pregnancy, injected antibody is unable to spread intensely throughout the brain—this is visible in Part A, the fourth and fifth images from the left, as dim patches within the skull of the embryonic mouse. Part B confirms this result quantitatively, showing a reduction in the concentration of antibody within the brain tissue of mice around the same timepoint, as the BBB forms and begins to prevent the antibody from reaching the embryonic brain. In contrast, the antibody spreads into the brains of germ-free embryonic mice at day 16.5 (Part C), suggesting that an effective BBB has not developed.

The authors confirm this result by slicing embryonic brains into thin sections and observing that the maternally injected antibody (Parts D and E, shown as red) is present more broadly in the germ-free embryonic brain. They labeled blood vessels (shown as green), and found that for pathogen-free mice, the antibody is present in high levels only within blood vessels. In contrast, the brains of germ-free mice exhibited strong antibody (red) signal throughout the brain slice.

Tight junctions

Having determined that embryonic germ-free mice lack an effective BBB, the authors wanted to identify the cause of the deficit. They tested the expression levels of three proteins that form tight junctions, which are critical components of the BBB.

Using Western blotting (see additional information in Figure 4, “Western blots”), they found that a specific protein, occludin, was less highly expressed in germ-free mice than pathogen-free ones (Parts F and G). The expression levels of two other proteins involved in tight junctions were unaffected, suggesting that occludin might be powerfully and specifically regulated by gut microbiota.

The next questions

These findings raise several questions:

1. All these observations were made in embryonic mice whose mothers were either germ-free or pathogen-free. At this point, it is impossible to know whether germ-free mice are truly unable to generate effective BBBs, or whether they simply lag behind their healthier pathogen-free counterparts, but could catch up later in development. Do germ-free mice continue to have defective BBBs into adulthood (eee Figure 2)?

2. The authors have shown that one component of tight junctions, occludin, is less highly expressed in the brains of embryonic germ-free mice. However, two others are expressed at normal levels. Does the loss of all gut microbiota truly only impact a single protein within the BBB? Are other components of the BBB affected by a germ-free environment?

Lack of gut microbiota is associated with increased BBB permeability in adult mice

Three techniques were used to determine whether the BBB was more permeable in germ-free adult mice: (i) in vivo positron emission tomography (PET) imaging with [11C]raclopride (Fig. 2, A to C); (ii) extravasation of Evans blue tracer from the circulation (Fig. 2D); and (iii) the capacity of an anti–N-methyl-D-aspartate receptor reactive antibody (R4A) to induce neuronal death after intravenous administration (Fig. 2, E and F).

Main question

Is the BBB in the adult germ-free mouse still more permeable than pathogen-free mice, or does the difference disappear as the mice age?

Experimental setup

The authors tested mice with three different compounds to measure BBB permeability. Each experiment involved a slightly different set of procedures and produced data of a different type, but in all cases, the authors introduced the compound into the circulatory system and later tested for its presence in the brain.

PET imaging

First, the authors injected mice intravenously with a radioactively labeled drug that is known to be able to pass through the BBBs of healthy mice. The radioactive tag on the drug allows the authors to use positron emission tomography to determine both the timing and degree of flux of the drug from the blood into the brains of germ-free and pathogen-free mice.

They find that although the drug can reach the brains of either type of mouse, it flows significantly more quickly into the brains of germ-free mice. This can be seen in Part A, which shows the influx of the drug in brains in the first 2 to 3 minutes following drug injection—note more red, orange, and yellow signal in the germ-free than the pathogen-free images. This difference is quantified in Parts B and C, which show increased uptake of the drug by the germ-free brains during the first 5 minutes postinjection.

These data suggest that the BBB in germ-free mice is more permeable than in pathogen-free mice.

Evans blue dye

Next, the authors combined dye injections and brain slicing to obtain higher resolution images of leakage into the brains.

They anesthetized mice and ran dye (shown as red in Part D) through the venous system of germ-free and pathogen-free mice. They thinly sliced the brains, then photographed them and determined the extent to which the dye had escaped the vasculature.

They found that germ-free mice showed large patches of dye spread through the tissue of several brain regions, whereas dye in pathogen-free mice was restricted to small lines likely corresponding only to the circulatory system (Part D, left and middle columns of images).

To verify that a permeable BBB would cause the types of dye patches they observed in the germ-free mice, they treated a group of pathogen-free mice with a protein (TNFα) that induces BBB leakage, then performed their dye injection and imaging experiments. They found very similar dye penetration into the brains of TNFα-treated mice as germ-free mice (Part D, right column), suggesting that the excessive dye observed in the brain tissue of germ-free mice can be explained by increased BBB permeability.

R4A antibody

Finally, the authors injected germ-free mice with either an antibody that causes neurons to die, or with harmless saline, then examined the brains 48 hours later.

Penetration of the antibody into the brain would be expected to cause cell death, and indeed, antibody treatment induced cell death in the hippocampus of germ-free mice (Parts E and F), whereas the brain neurons in the pathogen-free mice were unaffected (Part F).

This finding suggests that the BBB in the germ-free mice is permeable to large biological molecules capable of impairing neuronal health.

In germ-free adult mice, [11C]raclopride uptake was increased compared with that for pathogen-free adult mice (Fig. 2A), but only during the first 4 min after injection (Fig. 2, B and C). This period of time represents the “flow phase” (that is, the presence of the radioligand in the whole brain due to BBB permeability). Because the radioligand was given in tracer doses, it does not exert any pharmacological effects on the brain (or body) vasculature or heart rate. These differences were present only in the initial flow phase and not in the later phase of the time activity curves, indicating no differences in radioligand binding to dopamine D2 receptors between the groups.

Fluorescence microscopy images of different brain regions (cortex, striatum, and hippocampus) of pathogen-free adult mice showed the presence of Evans blue dye (bright red) only in the blood vessels, whereas Evans blue staining in germ-free mice was detected not only in the blood vessels but also in the brain parenchyma, demonstrating leakage of the dye across the BBB (Fig. 2D). A group of mice receiving an intravenous injection of tumor necrosis factor–α (TNFα) 15 hours before the experiment served as a positive control for BBB leakiness (Fig. 2D).

In germ-free adult mice, intravenous administration of the monoclonal antibody R4A (250 μg) was associated with abnormal neurons, marked by condensed cytoplasm and shrunken cell bodies in the CA1 region of the hippocampus (Fig. 2E, right panel). Abnormal neurons were not present or were rare in the CA1 region of the hippocampus in the control group [phosphate-buffered saline (PBS)–treated germ-free group; Fig. 2E, left panel]. Furthermore, the R4A injection (at low and high doses) in germ-free adult mice was associated with a significant reduction in neuron numbers in the CA1 region of the hippocampus, –38% for low-dose R4A (100 μg) and –42% for high-dose (250 μg) R4A compared with the PBS-treated germ-free control group (P < 0.01) (Fig. 2F). Intravenous administration of R4A in pathogen-free adult mice did not induce any changes, indicating that the R4A monoclonal antibody did not penetrate the BBB (Fig. 2F).

Vascular density and pericyte coverage show no difference in germ- and pathogen-free adult mice

We used intravital two-photon microscopy to exclude the possibility that high BBB permeability in germ-free adult mice was caused by higher vascular density in the brain. Using the second harmonic generation signals from collagen fibers of dura mater as a reference point, we observed that the subdural region, which is 20 to 80 μm below the dura mater, contains mainly large vessels (average diameter, ~40 μm) (Fig. 3, A and C). In contrast, deeper regions of the brain 120 to 180 μm below the dura mater consist mostly of capillaries (Fig. 3, B and D). Quantitative analysis of vasculature density in the brain revealed no significant gross differences between germ- and pathogen-free adult mice (Fig. 3, C and D).

Main question

Are there any obvious differences in the blood vessels of germ-free and pathogen-free mice?

The authors have shown that the blood vessels of germ-free mice are leakier than those of pathogen-free mice, and their results in Figure 1F suggest that the BBB is disrupted by the germ-free environment. However, the authors do not yet know whether blood vessels develop normally in germ-free mice. If the blood vessels were excessively large or dense, it would be difficult to conclude that the leakiness was due to the BBB, rather than the expanded presence of the blood vessels themselves.

The authors therefore examine the vascular system directly.

Experimental setup

The authors combine two imaging techniques, second harmonic imaging microscopy and two-photon microscopy, to examine blood vessels in the first 200 micrometers of the mouse brain. They inject a fluorescent dye into the blood, then detect the dye with a two-photon microscope to observe the size of the blood vessels. They measured blood vessel volume at two depths: 20 to 80 micrometers below the surface of the brain, where mostly large vessels reside, and 120 to 180 micrometers below the surface, which contains mostly capillaries.

Blood vessel volume

The authors found that blood vessels appear similar in germ-free and pathogen-free mice at both depths (Parts A and B). A volumetric analysis showed that blood vessels occupy the same proportion of the brain in both groups (Parts C and D). Thus, it is unlikely that the increased leakiness of blood vessels in germ-free mice is due to major changes in the shape or density of blood vessels.

Normal pericytes

Having concluded that the BBB is likely severely disrupted in germ-free mice, the authors begin to search for the specific abnormalities.

Pericytes are cells that wrap around the endothelial cells that line blood vessels to form the BBB. Pericytes make several contributions to the formation of the BBB: They help form tight junctions, they shield the developing BBB from damage by the immune system, and they help regulate the size of molecules that can pass through the BBB.

The authors use fluorescent antibodies to label the location of pericytes within the brain (Part E), and find that pericytes appear normal in germ-free mice.

Pericytes (green) can be observed running continuously along blood vessels (labeled red), in both the pathogen-free and germ-free groups. Thus, it appears unlikely that a deficit in the presence of pericytes explains the BBB leakiness of germ-free mice.

The next question

The authors have strong evidence that the BBB is disrupted in germ-free mice. However, one critical component of the BBB, pericytes, appears normal. What is the specific problem underlying the impaired BBB? See Figure 4.

Pericytes play an important role in regulating BBB properties, and decreased pericyte coverage has been associated with increased BBB permeability (10, 14). Immunofluorescence staining using CD13, a cell surface marker for pericytes in different brain regions, revealed no difference in pericyte coverage between germ- and pathogen-free mice (Fig. 3E).

Brain endothelial tight junctions are altered in the absence of a gut microbiota

Permeability of CNS vessels is controlled in part by dynamic opening and closing of the endothelial junctions (15). Therefore, we assessed the expression of the main tight junction proteins (ZO-1, occludin, and claudin-5) by Western blot in three regions of an adult mouse brain: frontal cortex, striatum, and hippocampus. Significantly lower expression of occludin and claudin-5 was observed in male germ-free mice compared with male pathogen-free mice in all three brain regions of interest (occludin: P < 0.001 in frontal cortex and hippocampus and P < 0.05 in striatum; claudin-5: P > 0.001 in frontal cortex and P < 0.01 in striatum and hippocampus) (Fig. 4, A to F). In contrast, no difference in the expression of the cytoplasmic protein ZO-1 was observed between the two groups (Fig. 4, A to F). Similar patterns of tight junction protein expression were observed in the brains of female germ-free mice compared with female pathogen-free mice (fig. S1), suggesting that the effect of gut microbiota on the integrity of the BBB is independent of sex. In addition, immunofluorescence analysis confirmed lower expression of occludin and claudin-5 in the brain vessels of adult male germ-free mice compared with that of adult male pathogen-free mice (Fig. 4, G and H).

Main questions

The authors suspected that the BBB is misdeveloped in germ-free mice. They had shown that one protein, occludin, is less highly expressed in embryonic germ-free mice than in pathogen-free mice, but so far in adult mice they had not found a mechanistic defect in the BBB. The authors next tested the expression levels of three proteins that form tight junctions, a critical component of the BBB, in adult mice. They examined three different brain areas.

Western blot

The authors tested the levels of three proteins: ZO-1, occludin, and claudin-5.

They used Western blotting, a technique in which tissue is ground up and dissolved using detergent. This process breaks apart cell membranes, releasing proteins. The proteins are labeled using antibodies that detect specific proteins with extremely high sensitivity, and then placed into an electronic device that draws them through a thick gel. The speed at which the proteins advance through the gel depends on their size, so this technique allows proteins to be separated so they can be detected individually.

The images in Parts A through C are photographs of these gels, and each dark band represents many copies of a single protein that has been isolated by being run through the gel, and labeled using an antibody. Proteins that are highly expressed appear as strong, black bands, whereas proteins with lower expression appear relatively faint. The authors show that in all three brain regions, occludin and claudin-5 are less highly expressed in germ-free mice than in pathogen-free mice (Parts A through C; compare the two leftmost bands [pathogen-free mice] with the two rightmost bands [germ-free mice] in each image).

The graphs in Parts D through F illustrate the quantitative comparisons of the intensity of the protein bands in germ-free and pathogen-free mice.

Fluorescent imaging

Parts G and H show that a reduction in occludin and claudin-5 in germ-free mice can be seen at blood vessels in the brain.

The authors generated brain slices, then labeled the two tight junction proteins with a fluorescent antibody (shown green in Parts G and H). Unlike Western blotting, this technique allowed the authors to determine whether occludin and claudin-5 was reduced specifically within blood vessels, rather than in other cell types within the brain.

The authors found that, indeed, claudin-5 and occludin were less highly expressed in the vasculature of germ-free mice (compare the brightness of the green signal in the left and right images with Parts G and H).

Electron microscopy

The authors have shown in Parts A through H that important tight junction proteins are expressed at reduced levels in the brains of germ-free mice, but still have not directly shown that tight junctions are disrupted.

In Parts I and J, they used transmission electron microscopy, an imaging technique that allows remarkably small structures within biological tissues to be directly observed, to assess tight junctions. They found that these junctions appear abnormal in germ-free mice: specifically, they find that they appear “blurred” in germ-free mice, rather than tight and linear (see additional examples in Supplemental Figure S2).

This result suggests that tight junctions are malformed in germ-free mice, which, in combination with their previous findings, is powerful evidence that gut microbes are necessary for the formation of tight junctions and a functional BBB.

The next questions

The authors have demonstrated that mice that are germ-free for their entire lives fail to develop normal tight junctions, and have abnormally permeable BBBs. An interesting question remains: Can the BBB be restored in adult mice by restoring their gut microbiota?

It would not be surprising if the BBB was irreparable in adult germ-free mice—many developmental processes, like the maturation of sensory systems and the acquisition of language, are difficult or impossible to undertake once development has ended. However, if the health of the BBB could be improved by normalization of the gut microbiota, this finding could potentially open new avenues of treatment for neurological diseases (see Figure 5).

Additionally, the authors have demonstrated that gut microbiota are important for the formation of the BBB, but have not established what specific molecules or signals are provided by the gut flora to promote BBB development. This information would add to our understanding of the interaction between the gut microbiota and the brain, and could help direct future clinical treatments (see Figure 6).

The ultrastructure of the tight junctions was investigated by transmission electron microscopy analysis. In germ-free adult mice, the tight junctions appeared as a diffuse, disorganized structure compared with those in the pathogen-free group (Fig. 4I). A scoring system was used to quantitatively determine the number of intact tight junctions as follows: perfect tight junctions, 3; patches of blurriness, 2; totally blurred, 1 (fig. S2 shows examples of the rating scale). In the striatum of germ-free adult mice, the number of intact tight junctions was significantly lower than that in pathogen-free mice (P < 0.001) (Fig. 4J).

BBB permeability and tight junction protein expression are associated with changes in the gut microbiota

Colonization of germ-free adult mice with flora from pathogen-free mice [conventionalized (CONV)] was associated with increased integrity of the BBB as shown by restriction of the Evans blue tracer to the blood vessels and decreased extravasation of the dye into the brain parenchyma (Fig. 5A). Quantitative analysis of tight junction proteins in the CONV group compared with germ-free mice showed a significant increase in the expression of occludin in the frontal cortex (P < 0.05) and striatum (P = 0.05) and of claudin-5 in the hippocampus and striatum (P < 0.05) (Fig. 5, B to G). Increased expression of the intracellular protein ZO-1 was detected in the striatum and hippocampus of CONV mice compared with germ-free controls (P < 0.05) (Fig. 5, B to G).

Main questions

The authors next attempted to restore BBB function by introducing a normal gut microbiota population into adult germ-free mice. If possible, this would suggest that certain neurological disorders could be treated through the improvement of gut microbiota health, a remarkable finding. They performed fecal transplantations from pathogen-free mice to germ-free mice, and after 2 weeks examined BBB permeability and the expression levels of the same proteins tested in Figure 4.

Permeability

The authors used transcardial injections of Evans blue dye, as in Figure 2, to test BBB permeability. They found that germ-free mice that had been “conventionalized” by the introduction of normal gut microbiota had BBBs that effectively prevented leakage of dye into the brain, unlike untreated germ-free mice. (Compare the large patches of red dye in the center column of Part A, representing germ-free mice, with the images of dye restricted to blood vessels in the right column of Part A, showing conventionalized mice).

Thus, the presence of normal gut microflora for only 2 weeks is sufficient to repair the damaged BBBs that develop in germ-free mice.

Protein levels

The authors had previously demonstrated in Figure 4 that two proteins, occludin and claudin-5, were expressed at lower levels in germ-free mice, whereas a third was unchanged.

Next, the authors tested the expression levels of these proteins in conventionalized mice, to determine whether gut microbiota had restored them to normal levels. They found increased expression of all three proteins in conventionalized mice (Parts B through G).

Note the large error bars in Parts E through G—conventionalization appears to increase protein levels to widely varying degrees. Because of this variability, expression of several of the proteins is not statistically significantly upregulated in some brain regions. However, in every case, expression of each protein at least trended towards an increase, suggesting that overall, conventionalization promoted the expression of tight junction proteins.

The next questions

The authors have shown that the introduction of normal gut microbiota into adult germ-free mice can restore the function of their BBBs after only 2 weeks. This striking result indicates that gut microbiota produce signals or proteins capable of reaching the brain and strongly promoting the development of the BBB.

What are these molecules, and how do they act in the brain (see Figure 6)?

In Parts I and J, they used transmission electron microscopy, an imaging technique that allows remarkably small structures within biological tissues to be directly observed, to assess tight junctions. They found that these junctions appear abnormal in germ-free mice: specifically, they find that they appear “blurred” in germ-free mice, rather than tight and linear (see additional examples in Supplemental Figure S2).

This result suggests that tight junctions are malformed in germ-free mice, which, in combination with their previous findings, is powerful evidence that gut microbes are necessary for the formation of tight junctions and a functional BBB.

SCFAs or metabolites produced by bacteria affect BBB permeability

SCFAs are known to enhance the integrity of the intestinal epithelial barrier (16, 17) by facilitating the assembly of tight junctions (18). Hence, we evaluated BBB permeability in germ-free adult mice monocolonized with a single bacterial strain, Clostridium tyrobutyricum (CBut), that produces mainly butyrate (19, 20) or with Bacteroides thetaiotaomicron (BTeta), which produces mainly acetate and propionate (21, 22). We also evaluated germ-free adult mice given sodium butyrate by oral gavage for 3 days. Evans blue perfusion in CBut-, BTeta-, and sodium butyrate–treated mice demonstrated decreased BBB permeability, compared to that in germ-free adult mice, that was equivalent to that of pathogen-free adult mice (Fig. 6A). Administration of sodium butyrate to germ-free mice was associated with increased expression of occludin in the frontal cortex and hippocampus but had no effect on the expression of claudin-5 (Fig. 6, B to D). Furthermore, administration of sodium butyrate or monocolonization of germ-free mice with C. tyrobutyricum was associated with an increase in histone acetylation in brain lysates (fig. S3).

Main questions

The authors showed that restoring the gut microbiota rapidly rescues normal functionality to the BBBs of germ-free mice. Next, they investigated a type of molecule that is commonly produced by bacteria as a metabolic byproduct: SCFAs. SCFAs have previously been shown to promote tight junction formation between endothelial cells in the intestine.

Could the gut microbiome produce SCFAs that reached the brain and assisted in the development of the BBB?

Experimental setup

The authors colonized the digestive tracts of germ-free mice with one of two species of bacteria, each of which produces mainly SCFAs (labeled CBut and BTeta in Figure 6). Additionally, they delivered a purified SCFA to some germ-free mice (labeled NaBu).

Each of these treatments was designed to elevate the level of SCFAs in germ-free mice. After treatment, the authors tested BBB permeability by injecting Evans blue dye as in Figure 2.

They also examined the expression levels of core tight junction proteins using Western blots, as in Figure 4.

Permeability

After various treatments to increase SCFA levels in germ-free mice, the authors tested BBB permeability by injecting the mice with Evans blue dye, as in Figure 2. They found that in each treatment group, the BBBs of treated mice were as effective as those in normal pathogen-free mice (Part A). This indicates that even a short treatment with one or two SCFAs is sufficient to substantially restore the performance of the damaged BBBs in germ-free mice.

Protein levels

Next, the authors asked whether treatment with a purified SCFA restored the expression levels of the core components of tight junctions. Using Western blots, as in Figure 2, they found that the expression of one protein, occludin, was increased in two brain regions by SCFA treatment, though it was unaffected in a third region. Other tight junction proteins were not changed by treatment, suggesting that SCFAs may specifically regulate the level of occludin.

DISCUSSION

The BBB is a physiological barrier that controls the passage of molecules between the brain parenchyma and the blood and in so doing allows proper functioning of neurons. Our results highlight the gut microbiota as a potential regulator of BBB integrity. Here, we show that the lack of a normal gut microbiota in germ-free mice is associated with increased permeability of the BBB. This result was confirmed using three different techniques: in vivo PET imaging using radiolabeled ligand, vascular leakage of Evans blue dye, and neuronal damage after intravenous administration of R4A antibody. Furthermore, our data show that a more permeable BBB is observed in the fetal mice with germ-free mothers at E16.5 to E18.5 days of embryonic development compared to the fetal mice with pathogen-free mothers at the same stages of development. The increased permeability of the BBB in germ-free adult mice may partly be the consequence of disorganized tight junctions, as shown by electron microscopy analysis, and low expression of the transmembrane tight junction proteins occludin and claudin-5. The “conventionalization” of germ-free adult mice through transplant of the fecal microbiota from pathogen-free adult mice or by administering bacterial strains that produce SCFAs reinforced the integrity of the mouse BBB.

The BBB matures progressively during intrauterine life and continues to mature during the early postnatal stages of life (23). Our data confirm previous observations (10, 24) and show that closure of the BBB to IgG in pathogen-free mice occurs during the later stages of intrauterine life. A recent study of Mfsd2a knockout mice (lacking the transporter for the essential omega-3 fatty acid docosahexaenoic acid) showed that the BBB in mice becomes functional at E15.5, demonstrating complex regional and temporal differences in maturation (11). This coincides with our observation of the permeability of the embryonic BBB to maternal antibodies. In mice, gestational stage E15.5 is a turning point for the restriction of maternal antibody penetration into the fetal brain. Maternal antibodies or, more precisely, antibody delivered to the embryo through the placenta was our molecule of choice in our embryonic BBB studies as a physiological route of delivery. Reduced closure of the BBB was observed in fetuses from germ-free dams. In humans, marked changes in the composition of the maternal gut microbiota have been observed between the first and the third trimesters of pregnancy (25). These observations, together with the present study, imply that the maternal gut microbiota might contribute to increased nutritional demands in late pregnancy, which would require more stringent control of BBB permeability in the growing offspring.

The BBB is a complex structure formed by capillary endothelial cells, pericytes, and astrocytes (26). A difference in vascular density might be a confounding factor in assessing BBB permeability. In our study, increased invasion of circulating substances into the brain parenchyma appears not to be due to differences in large vascular structures between the two groups of mice as shown by a comparable equal visualization of the brain vasculature using 140 kD TRITC-dextran. However, we cannot formally exclude that some microcapillary changes may still exist between the two groups. In germ-free adult mice, pericyte coverage was similar to that in pathogen-free mice, suggesting that altering the number of pericytes is unlikely to account for the increased BBB permeability observed in germ-free mice. This is similar to the Mfsd2a-deficient mouse model in which leaky vessels are not associated with changes in the cerebrovascular network or pericyte abnormalities (11). However, in contrast to the Mfsd2a-deficient mouse where transcytosis across the BBB is affected without disruption of tight junctions, our model implies that the gut microbiota may regulate the BBB through modulation of tight junction protein expression.

Tight junctions play a major role in the functional maintenance of BBB (27, 28). Our data show disorganized tight junctions in the brains of germ-free adult mice compared with those of pathogen-free mice, which was associated with lower expression of occludin and claudin-5. No difference in the expression of ZO-1 was observed. A decrease in occludin and claudin-5 paralleled by cytoskeletal changes and tight junction protein redistribution was associated with altered integrity of the BBB (29). Reduced expression of occludin was observed in germ-free mice during both intrauterine life and adulthood and was associated with increased permeability of the BBB. Administration of normal flora from pathogen-free mice or oral treatment with the bacterial metabolite sodium butyrate to germ-free adult mice induced an increase in the expression of occludin that was associated with decreased permeability of the BBB. These observations imply that the expression of occludin by the brain endothelial cells is sensitive to changes in the intestinal gut microbiota.

Regulation of occludin expression by the intestinal microbiota has been reported in the intestinal epithelial barrier (30) and blood-testis barrier (31). The expression of claudin-5 by brain endothelial cells was low in germ-free adult mice but elevated after exposure to the gut microbiota. However, no difference in the expression of claudin-5 was observed in the E18.5 cortex of fetal mice with germ- or pathogen-free mothers. This implies that claudin-5 expression continues to increase during the postnatal period in offspring of pathogen-free mothers but not in offspring of germ-free mice. Although claudin-5 expression is present from early intrauterine life in the human brain, it is localized mostly in the endothelial cytoplasm and shifts toward the endothelial border during development to form tight junctions (32, 33). Furthermore, claudin-5 knockout mice appear to develop normally during intrauterine life with morphologically normal blood vessels but have a BBB impairment associated with increased permeability to small molecules (34). The complete depletion of claudin-5 is associated with 100% mortality of the newborns within 10 hours of birth (34), further supporting the role of claudin-5 in the postnatal regulation of the BBB.

Dietary carbohydrates are substrates for fermentation by certain gut bacteria, which produce SCFAs, primarily acetate, propionate, and butyrate, as end products (35). These SCFAs have been shown to regulate intestinal motility (36, 37) and to be involved in central appetite regulation (38, 39) as well as being taken up directly into the bloodstream and transported to various organs, including the brain (39, 40), where they modulate tissue development and function (41,42). Physiological concentrations of SCFA mixtures or individual SCFAs regulate intestinal barrier function by increasing the transepithelial electrical resistance and decreasing paracellular permeability (43). In a rat model of transient focal cerebral ischemia, intraperitoneally injected sodium butyrate attenuated BBB disruption (6). In our study, we show that monocolonization of the intestine of germ-free adult mice with either C. tyrobutyricum, a bacterial strain producing butyrate (19, 44), or B. thetaiotaomicron, which produces mainly acetate and propionate (21), decreased BBB permeability. Oral administration of the bacterial metabolite sodium butyrate mimicked this effect on the BBB. The effects of the other metabolites, acetate and propionate, as single substrates may also have an effect on the permeability of the BBB and should also be explored. Intravenous or intraperitoneal administration of sodium butyrate has been reported to inhibit histone deacetylation and facilitate long-term memory consolidation (45), prevent BBB breakdown (46), and promote angiogenesis and neurogenesis (47,48). Whether the gut microbiota and sodium butyrate alter histone acetylation of brain microvascular endothelial cells requires further study to better understand the effects of SCFAs produced by the gut microbiota on the CNS. Cumulative and chronic exposure to SCFAs could lead to relatively stable effects on gene expression in the brain.

Finally, the composition and diversity of the gut microbiota community change over time, presumably reflecting different gut microbiota–host interactions. Early in life, there is an urgent need to support the growing offspring with an almost unlimited amount of energy to ensure brain development. Germ-free mice show an increase in glucocorticoid production due to metabolic stress (49). An increase in BBB permeability may, therefore, allow serum glucocorticoids to enter the growing brain, affecting neurogenesis and impairing production of brain-derived nerve growth factor in the brain (6). In addition, microbes regulating the BBB in infants with a growing brain may influence BBB permeability differently compared to the adult BBB (50).

Although providing a first glimpse into an additional layer of gut-brain communication, the current study has not revealed the precise signaling mechanisms through which gut microbiota modulates BBB function. In addition, the consequences of increased BBB permeability in germ-free mice on neuronal function throughout life are still not known. Therefore, the results should be interpreted cautiously until more physiological data are acquired to corroborate these findings.

MATERIALS AND METHODS

Study design

The objective of this study was to assess the importance of the intestinal microbiota in the maintenance of BBB integrity in a mouse model. The integrity of the BBB was examined in germ-free mice and in pathogen-free mice using functional permeability assays and by determining the status of tight junctions. BBB integrity was also determined in a group of germ-free adult mice colonized with fecal samples from pathogen-free adult mice or treated with bacterial strains that produced SCFAs or the bacterial metabolite sodium butyrate.

Animals

Germ- and pathogen-free NMRI (Naval Medical Research Institute) male mice and C57BL/6J female and male mice (8 to 10 weeks old) (Core Facility for Germ-free Research, Karolinska Institutet) were used. C57BL/6J, Balb/c, and NMRI germ- and pathogen-free female mice undergoing timed pregnancies were used to assess the intrauterine development of the BBB. Germ-free mice were raised in special plastic isolators. All animals were maintained on autoclaved R36 Lactamin chow (Lactamin), given sterile drinking water ad libitum, and kept under 12-hour light/dark cycles.

The protocols were approved by the Regional Animal Research Ethical Board, Stockholm, Sweden, following the proceedings described in the (European Union) EU legislation (Council Directive 2010/63/EU) and the Institutional Animal Care and Use Committee at The Feinstein Institute for Medical Research, Manhasset, NY, USA.

Colonization of germ-free mice

For colonization, 8- to 10-week-old C57BL/6J germ-free mice received fecal matter from the pathogen-free mice through a single gavage and then were left for 14 days (CONV) before being sacrificed. The control group was gavaged with sterile PBS.

C. tyrobutyricum (DSM 2637) (a contribution of H. Kozakova, Department of Immunology and Gnotobiology, Institute of Microbiology, Academy of Sciences of the Czech Republic, Praha, Czech Republic) and B. thetaiotaomicron (BTeta) were cultured in Bryant Burkey broth with resazurine (Merck), and 108 bacteria were used to colonize each C57BL/6J germ-free adult male mouse as previously described (19). Fourteen days later, the mice were treated with Evans blue dye to assess the BBB permeability. Another group of germ-free male mice was gavaged for 3 days with sodium butyrate [1 g/kg body weight (BW) per day] before sacrifice. A group of germ-free male mice and another group of pathogen-free mice were gavaged with sterile water and used as controls.

Drug treatment

A group of adult male pathogen-free mice receiving a single intravenous injection of TNFα (100 μg/kg BW) 15 hours before Evans blue perfusion were used as a positive control for a “leaky” BBB (51).

Perfusion with Evans blue dye

Evans blue perfusion was performed as previously described (52). Briefly, anesthetized mice were perfused with PBS followed by Evans blue cocktail. Tissue cryosections were analyzed by fluorescence microscopy. A detailed description is provided in Supplementary Materials.

In vivo PET imaging of [11C]raclopride

Scans were performed in a microPET Focus 120 scanner (CTI Concorde Microsystems). The tracer (maximum volume, 200 μl) was administered by bolus injection via the tail vein into anesthetized mice. PET data were acquired in full three-dimensional mode, and images were reconstructed by standard two-dimensional filtered back projection using a ramp filter. The % SUV is the regional concentration of tissue radioactivity normalized for injected dose and BW. A complete description is provided in Supplementary Materials.

R4A antibody injection and quantitative analysis

Germ- and pathogen-free adult male mice received intravenous injections of R4A dissolved in PBS. Forty-eight hours later, anesthetized mice were given transcardiac perfusion with 0.9% NaCl followed by 4% paraformaldehyde (PFA) (Histolab). Tissue cryosections stained with cresyl violet and neurons were sampled from comparable regions of the anterior dorsal hippocampus (53,54). A detailed protocol is provided in Supplementary Materials.

Intravital two-photon laser scanning microscopy

TRITC-dextran (155 kD, Sigma) was injected retro-orbitally before imaging. Two-photon imaging was performed on a TriM Scope II (LaVision BioTec) equipped with an Olympus BX51 upright microscope fitted with a 20 × 0.95 numerical aperture water immersion objective and a Chameleon Ultra-II Tunable, Mode-locked Ti:Sapphire laser (Coherent) tuned to 880 nm for excitation of TRITC-dextran. Imaris (Bitplane) was used for three-dimensional image analysis. Volumetric index (VI), a parameter that indicates percentage of volume occupied by blood vessels within the analyzedZ stack, was calculated according to the following formula: VI = total blood vessel volume within the Z stack/total volume of the Z stack × 100.

Immunofluorescence

Immunofluorescence was performed on 50-μm brain coronal cryotome sections blocked with 5% bovine serum albumin (BSA) (Sigma) and 0.5% Triton X-100 (Sigma) in PBS, followed by incubation with primary antibody as follows: rat anti-CD13 (BD Pharmingen, 558744), mouse anti-occludin (1:300, Invitrogen, 33-1500), mouse anti–claudin-5 (1:300, Invitrogen, 35-2500), and rabbit anti-laminin (1:400, Sigma, L9393). Sections were then incubated with secondary donkey anti-rat Alexa 488, goat anti-mouse Alexa 488, or goat anti-rabbit Alexa 594 (1:500, Invitrogen) for 1 hour at room temperature. Z stack images were visualized and acquired using a Nikon Eclipse TE300 inverted microscope integrated with a PerkinElmer UltraVIEW spinning disk confocal system. Image processing was done using Corel Paint Shop Pro Photo XI software.

Protein extraction and Western blot analysis

Protein expression was determined in brain lysates as previously described (19). Primary antibodies (occludin: 1:1000, Invitrogen, 331500; claudin-5: 1:1000, Invitrogen, 352500; and ZO-1: 1:2000, Invitrogen, 617300) diluted in 3% BSA in 0.1% Tween 20 (Sigma)/PBS were incubated overnight at 4°C with the blot. Protein bands were visualized by chemiluminescence using Immun-Star WesternC chemiluminescence kit (Bio-Rad). Protein expression was then quantified using ImageJ software (National Institute of Mental Health).

Transmission electron microscopy

A group of adult pathogen- and germ-free mice were transcardially perfused with 2.5% glutaraldehyde (Sigma)/1% PFA and processed for transmission electron microscopy. Digital images were taken with a Veleta camera (Olympus Soft Imaging Solutions, GmbH). The full method is described in Supplementary Materials.

Statistical analysis

Statistical significance was determined using one-way ANOVA with Tukey post hoc test for multiple groups, Bonferroni post hoc analysis for Fig. 1B, or t test when changes were compared between two groups (GraphPad Prism 6). P < 0.05 was considered statistically significant unless otherwise stated. Values were expressed as means ± SEM.

Supplementary Materials

www.sciencetranslationalmedicine.org/cgi/content/full/6/263/263ra158/DC1

Material and Methods
Fig. S1. Expression of tight junction proteins in the brains of germ- and pathogen-free adult female mice.
Fig. S2. Electron micrographs showing different tight junction structure (white arrows) in the brains of germ-free adult mice.
Fig. S3. The effect of oral treatment with the bacterial metabolite sodium butyrate or monocolonization with C. tyrobutyricum on histone acetylation in extracts of mouse brain frontal cortex.

References and Notes

1. F. Bäckhed, R. E. Ley, J. L. Sonnenburg, D. A. Peterson, J. I. Gordon, Host-bacterial mutualism in the human intestine. Science 307, 1915–1920 (2005).

2. L. V. Hooper, Bacterial contributions to mammalian gut development. Trends Microbiol. 12, 129–134 (2004).

3. V. R. Velagapudi, R. Hezaveh, C. S. Reigstad, P. Gopalacharyulu, L. Yetukuri, S. Islam, J. Felin, R. Perkins, J. Borén, M. Oresic, F. Bäckhed, The gut microbiota modulates host energy and lipid metabolism in mice. J. Lipid Res. 51, 1101–1112 (2010).

4. J. K. Nicholson, E. Holmes, J. Kinross, R. Burcelin, G. Gibson, W. Jia, S. Pettersson, Host-gut microbiota metabolic interactions. Science 336, 1262–1267 (2012).

5. L. V. Hooper, D. R. Littman, A. J. Macpherson, Interactions between the microbiota and the immune system. Science 336, 1268–1273 (2012).

6. R. Diaz Heijtz, S. Wang, F. Anuar, Y. Qian, B. Björkholm, A. Samuelsson, M. L. Hibberd, H. Forssberg, S. Pettersson, Normal gut microbiota modulates brain development and behavior. Proc. Natl. Acad. Sci. U.S.A. 108, 3047–3052 (2011).

7. B. Engelhardt, Development of the blood-brain barrier. Cell Tissue Res. 314, 119–129 (2003).

8. T. K. Hensch, Critical period regulation. Annu. Rev. Neurosci. 27, 549–579 (2004).

9. E. I. Knudsen, Sensitive periods in the development of the brain and behavior. J. Cogn. Neurosci. 16, 1412–1425 (2004).

10. R. Daneman, L. Zhou, A. A. Kebede, B. A. Barres, Pericytes are required for blood–brain barrier integrity during embryogenesis. Nature 468, 562–566 (2010).

11. A. Ben-Zvi, B. Lacoste, E. Kur, B. J. Andreone, Y. Mayshar, H. Yan, C. Gu, Mfsd2a is critical for the formation and function of the blood–brain barrier. Nature 509, 507–511 (2014).

12. B. T. Hawkins, T. P. Davis, The blood-brain barrier/neurovascular unit in health and disease. Pharmacol. Rev. 57, 173–185 (2005).

13. C. Tscheik, I. E. Blasig, L. Winkler, Trends in drug delivery through tissue barriers containing tight junctions. Tissue Barriers 1, e24565 (2013).

14. A. Armulik,G. Genové, M. Mäe, M.H. Nisancioglu, E. Wallgard, C. Niaudet, L. He, J. Norlin, P. Lindblom, K. Strittmatter, B. R. Johansson, C. Betsholtz, Pericytes regulate the blood–brain barrier. Nature 468, 557–561 (2010).

15. E. Dejana, E. Tournier-Lasserve, B. M. Weinstein, The control of vascular integrity by endothelial cell junctions: Molecular basis and pathological implications. Dev. Cell 16, 209–221 (2009).

16. J. M. Mariadason, D. H. Barkla, P. R. Gibson, Effect of short-chain fatty acids on paracellular permeability in Caco-2 intestinal epithelium model. Am. J. Physiol. 272, G705–G712 (1997).

17. J. M. Mariadason, A. Catto-Smith, P. R. Gibson, Modulation of distal colonic epithelial barrier function by dietary fibre in normal rats. Gut 44, 394–399 (1999).

18. L. Peng, Z. He, W. Chen, I. R. Holzman, J. Lin, Effects of butyrate on intestinal barrier function in a Caco-2 cell monolayer model of intestinal barrier. Pediatr. Res. 61, 37–41 (2007).

19. A. Korecka, T. de Wouters, A. Cultrone, N. Lapaque, S. Pettersson, J. Doré, H. M. Blottière, V. Arulampalam, ANGPTL4 expression induced by butyrate and rosiglitazone in human intestinal epithelial cells utilizes independent pathways. Am. J. Physiol. Gastrointest. Liver Physiol. 304, G1025–G1037 (2013).

20. P. Louis, H. J. Flint, Diversity, metabolism and microbial ecology of butyrate-producing bacteria from the human large intestine. FEMS Microbiol. Lett. 294, 1–8 (2009).

21. J. Noack, G. Dongowski, L. Hartmann, M. Blaut, The human gut bacteria Bacteroides thetaiotaomicron and Fusobacterium varium produce putrescine and spermidine in cecum of pectin-fed gnotobiotic rats. J. Nutr. 130, 1225–1231 (2000).

22. B. S. Samuel, J. I. Gordon, A humanized gnotobiotic mouse model of host–archaeal–bacterial mutualism. Proc. Natl. Acad. Sci. U.S.A. 103, 10011–10016 (2006).

23. U. Kniesel, W. Risau, H. Wolburg, Development of blood-brain barrier tight junctions in the rat cortex. Brain Res. Dev. Brain Res. 96, 229–240 (1996).

24. J. A. Siegenthaler, F. Sohet, R. Daneman, ‘Sealing off the CNS’: Cellular and molecular regulation of blood–brain barriergenesis. Curr. Opin. Neurobiol. 23, 1057–1064 (2013).

25. O. Koren, J. K. Goodrich, T. C. Cullender, A. Spor, K. Laitinen, H. K. Bäckhed, A. Gonzalez, J. J. Werner, L. T. Angenent, R. Knight, F. Bäckhed, E. Isolauri, S. Salminen, R. E. Ley, Host remodeling of the gut microbiome and metabolic changes during pregnancy. Cell 150, 470–480 (2012).

26. N. J. Abbott, A. A. Patabendige, D. E. Dolman, S. R. Yusof, D. J. Begley, Structure and function of the blood-brain barrier. Neurobiol. Dis. 37, 13–25 (2010).

27. H. Jiao, Z. Wang, Y. Liu, P. Wang, Y. Xue, Specific role of tight junction proteins claudin-5, occludin, and ZO-1 of the blood–brain barrier in a focal cerebral ischemic insult. J. Mol. Neurosci. 44, 130–139 (2011).

28. H. Wolburg, A. Lippoldt, Tight junctions of the blood–brain barrier: Development, composition and regulation. Vascul. Pharmacol. 38, 323–337 (2002).

29. G. Schreibelt, G. Kooij, A. Reijerkerk, R. van Doorn, S. I. Gringhuis, S. van der Pol, B. B. Weksler, I. A. Romero, P. O. Couraud, J. Piontek, I. E. Blasig, C. D. Dijkstra, E. Ronken, H. E. de Vries, Reactive oxygen species alter brain endothelial tight junction dynamics via RhoA, PI3 kinase, and PKB signaling. FASEB J. 21, 3666–3676 (2007).

30. P. D. Cani, S. Possemiers, T. Van de Wiele, Y. Guiot, A. Everard, O. Rottier, L. Geurts, D. Naslain, A. Neyrinck, D. M. Lambert, G. G. Muccioli, N. M. Delzenne, Changes in gut microbiota control inflammation in obese mice through a mechanism involving GLP-2-driven improvement of gut permeability. Gut 58, 1091–1103 (2009).

31. M. Al-Asmakh, J. B. Stukenborg, A. Reda, F. Anuar, M. L. Strand, L. Hedin, S. Pettersson, O. Söder, The gut microbiota and developmental programming of the testis in mice. PLOS One 9, e103809 (2014).

32. J. A. Anstrom, C. R. Thore, D. M. Moody, W. R. Brown, Immunolocalization of tight junction proteins in blood vessels in human germinal matrix and cortex. Histochem. Cell Biol. 127, 205–213 (2007).

33. D. Virgintino, M. Errede, D. Robertson, C. Capobianco, F. Girolamo, A. Vimercati, M. Bertossi, L. Roncali, Immunolocalization of tight junction proteins in the adult and developing human brain. Histochem. Cell Biol. 122, 51–59 (2004).

34. T. Nitta, M. Hata, S. Gotoh, Y. Seo, H. Sasaki, N. Hashimoto, M. Furuse, S. Tsukita, Size-selective loosening of the blood-brain barrier in claudin-5–deficient mice. J. Cell Biol. 161, 653–660 (2003).

35. S. E. Pryde, S. H. Duncan, G. L. Hold, C. S. Stewart, H. J. Flint, The microbiology of butyrate formation in the human colon. FEMS Microbiol. Lett. 217, 133–139 (2002).

36. P. S. Kamath, M. T. Hoepfner, S. F. Phillips, Short-chain fatty acids stimulate motility of the canine ileum. Am. J. Physiol. 253, G427–G433 (1987).

37. A. Richardson, A. T. Delbridge, N. J. Brown, R. D. Rumsey, N. W. Read, Short chain fatty acids in the terminal ileum accelerate stomach to caecum transit time in the rat. Gut 32, 266–269 (1991).

38. M. L. Sleeth, E. L. Thompson, H. E. Ford, S. E. Zac-Varghese, G. Frost, Free fatty acid receptor 2 and nutrient sensing: A proposed role for fibre, fermentable carbohydrates and shortchain fatty acids in appetite regulation. Nutr. Res. Rev. 23, 135–145 (2010).

39. G. Frost, M. L. Sleeth, M. Sahuri-Arisoylu, B. Lizarbe, S. Cerdan, L. Brody, J. Anastasovska, S. Ghourab, M. Hankir, S. Zhang, D. Carling, J. R. Swann, G. Gibson, A. Viardot, D. Morrison, E. Louise Thomas, J. D. Bell, The short-chain fatty acid acetate reduces appetite via a central homeostatic mechanism. Nat. Commun. 5, 3611 (2014).

40. F. De Vadder, P. Kovatcheva-Datchary, D. Goncalves, J. Vinera, C. Zitoun, A. Duchampt, F. Bäckhed, G. Mithieux, Microbiota-generated metabolites promote metabolic benefits via gut-brain neural circuits. Cell 156, 84–96 (2014).

41. W. Scheppach, Effects of short chain fatty acids on gut morphology and function. Gut 35, S35–S38 (1994).

42. D. F. Macfabe, Short-chain fatty acid fermentation products of the gut microbiome: Implications in autism spectrum disorders. Microb. Ecol. Health Dis. 23, 19260 (2012).

43. T. Suzuki, S. Yoshida, H. Hara, Physiological concentrations of short-chain fatty acids immediately suppress colonic epithelial permeability. Br. J. Nutr. 100, 297–305 (2008).

44. T. Hudcovic, J. Kolinska, J. Klepetar, R. Stepankova, T. Rezanka, D. Srutkova, M. Schwarzer, V. Erban, Z. Du, J. M. Wells, T. Hrncir, H. Tlaskalova-Hogenova, H. Kozakova, Protective effect of Clostridium tyrobutyricum in acute dextran sodium sulphate-induced colitis: Differential regulation of tumour necrosis factor-a and interleukin-18 in BALB/c and severe combined immunodeficiency mice. Clin. Exp. Immunol. 167, 356–365 (2012).

45. J. M. Levenson, K. J. O’Riordan, K. D. Brown, M. A. Trinh, D. L. Molfese, J. D. Sweatt, Regulation of histone acetylation during memory formation in the hippocampus. J. Biol. Chem. 279, 40545–40559 (2004).

46. E. B. Fessler, F. L. Chibane, Z. Wang, D. M. Chuang, Potential roles of HDAC inhibitors in mitigating ischemia-induced brain damage and facilitating endogenous regeneration and recovery. Curr. Pharm. Des. 19, 5105–5120 (2013).

47. H. J. Kim, P. Leeds, D. M. Chuang, The HDAC inhibitor, sodium butyrate, stimulates neurogenesis in the ischemic brain. J. Neurochem. 110, 1226–1240 (2009).

48. D. Y. Yoo, W. Kim, S. M. Nam, D. W. Kim, J. Y. Chung, S. Y. Choi, Y. S. Yoon, M. H. Won, I. K. Hwang, Synergistic effects of sodium butyrate, a histone deacetylase inhibitor, on increase of neurogenesis induced by pyridoxine and increase of neural proliferation in the mouse dentate gyrus. Neurochem. Res. 36, 1850–1857 (2011).

49. N. Sudo, Y. Chida, Y. Aiba, J. Sonoda, N. Oyama, X. N. Yu, C. Kubo, Y. Koga, Postnatal microbial colonization programs the hypothalamic–pituitary–adrenal system for stress response in mice. J. Physiol. 558, 263–275 (2004).

50. M. J. Claesson, I. B. Jeffery, S. Conde, S. E. Power, E. M. O’Connor, S. Cusack, H. M. Harris, M. Coakley, B. Lakshminarayanan, O. O’Sullivan, G. F. Fitzgerald, J. Deane, M. O’Connor, N. Harnedy, K. O’Connor, D. O’Mahony, D. van Sinderen, M. Wallace, L. Brennan, C. Stanton, J. R. Marchesi, A. P. Fitzgerald, F. Shanahan, C. Hill, R. P. Ross, P. W. O’Toole, Gut microbiota composition correlates with diet and health in the elderly. Nature 488, 178–184 (2012).

51. M. Tsuge, K. Yasui, T. Ichiyawa, Y. Saito, Y. Nagaoka, M. Yashiro, N. Yamashita, T. Morishima, Increase of tumor necrosis factor-a in the blood induces early activation of matrix metalloproteinase- 9 in the brain. Microbiol. Immunol. 54, 417–424 (2010).

52. J. del Valle, A. Camins, M. Pallàs, J. Vilaplana, C. Pelegrí, A new method for determining blood–brain barrier integrity based on intracardiac perfusion of an Evans Blue–Hoechst cocktail. J. Neurosci. Methods 174, 42–49 (2008).

53. C. Kowal, L. A. DeGiorgio, T. Nakaoka, H. Hetherington, P. T. Huerta, B. Diamond, B. T. Volpe, Cognition and immunity: Antibody impairs memory. Immunity 21, 179–188 (2004).

54. B. T. Volpe, J. Wildmann, C. A. Altar, Brain-derived neurotrophic factor prevents the loss of nigral neurons induced by excitotoxic striatal-pallidal lesions. Neuroscience 83, 741–748 (1998).