Suppression of gut dysbiosis by Bifidobacterium longum alleviates cognitive decline in 5XFAD transgenic and aged mice

To understand the role of commensal gut bacteria on the progression of cognitive decline in Alzheimer’s disease via the microbiota-gut-brain axis, we isolated anti-inflammatory Bifidobacterium longum (NK46) from human gut microbiota, which potently inhibited gut microbiota endotoxin production and suppressed NF-κB activation in lipopolysaccharide (LPS)-stimulated BV-2 cells, and examined whether NK46 could simultaneously alleviate gut dysbiosis and cognitive decline in male 5xFAD-transgenic (5XFAD-Tg, 6 months-old) and aged (18 months-old) mice. Oral administration of NK46 (1 × 109 CFU/mouse/day for 1 and 2 months in aged and Tg mice, respectively) shifted gut microbiota composition, particularly Proteobacteria, reduced fecal and blood LPS levels, suppressed NF-κB activation and TNF-α expression, and increased tight junction protein expression in the colon of 5XFAD-Tg and aged mice. NK46 treatment also alleviated cognitive decline in 5XFAD-Tg and aged mice. Furthermore, NK46 treatment suppressed amyloid-β, β/γ-secretases, and caspase-3 expression and amyloid-β accumulation in the hippocampus of 5XFAD-Tg mice. NK46 treatment also reduced Iba1+, LPS+/CD11b+, and caspase-3+/NeuN+ cell populations and suppressed NF-κB activation in the hippocampus of 5XFAD-Tg and aged mice, while BDNF expression was increased. These findings suggest that the suppression of gut dysbiosis and LPS production by NK46 can mitigate cognitive decline through the regulation of microbiota LPS-mediated NF-κB activation.


Results
Bifidobacterium longum suppressed gut microbiota LpS production and LpS-induced nf-κB activation in BV-2 cells. We examined the inhibitory effects of 25 Lactobacilli and 25 Bifidobacteria strains on gut microbiota LPS production. Of these, NK46 most potently inhibited LPS production and fecal bacterial growth (Fig. 1A,B). NK46 also potently inhibited LPS-induced NF-κB activation in microglial BV-2 cells (Fig. 1C). NK46 was identified as Bifidobacterium longum, based on the results of Gram staining, API 20A Kit (bioMerieux, Seoul, Korea), and 16S rDNA sequencing (ABI 3730XL DNA analyzer, Thermo Fisher Scientific Inc, Waltham, MA, USA). NK46 shifted gut microbiota composition and suppressed gut microbiota LPS production and colitis marker expression in 5XFAD-Tg mice. To examine whether NK46 could fine-tune gut microbiota composition, we orally gavaged NK46 in 5XFAD-Tg mice and measured the fecal microbiota composition BV-2 cells (5 × 10 5 cells/mL) were incubated with LPS in the absence or presence of NK46 (1 × 10 5 CFU/mL). All data were expressed as mean ± SD (n = 4). * p < 0.05 vs. control group treated with vehicle alone.
using pyrosequencing (Fig. 2). Bacterial richness and α-diversity were significantly lower in 5XFAD-Tg mice than in control mice, as demonstrated by the number of sequences analyzed, estimated operational taxonomic unit (OTU) richness, abundance-based coverage estimator (ACE), Chao1, Shannon, and Simpson. (Fig. 2A). However, NK46 treatment increased bacterial richness and α-diversity. Comparing the results of taxonomy-based analysis between 5XFAD-Tg mice treated and not treated with NK46, it showed that the fecal microbiota composition of 5XFAD-Tg mice was significantly different from that of NK46-treated mice (Fig. 2B). At the phylum level, 5XFAD-Tg mice exhibited Proteobacteria and Firmicutes populations more abundantly than control mice, as previously reported 19 , while the Bacteroidetes population was lower in 5XFAD-Tg mice. Treatment with NK46 in 5XFAD-Tg mice decreased the populations of Firmicutes and Proteobacteria and increased the population of Bacteroidetes. At the class level, 5XFAD-Tg mice exhibited Clostridia, and δ-, γ-, and ε-Proteobacteria populations more abundantly than control mice, while Bacteroidia and Bacilli populations were lower. Treatment with NK46 in 5XFAD-Tg mice increased Bacteroidia population and suppressed Clostridia and δ-, γ-, and ε-Proteobacteria populations. At the family level, Lachnospiraceae, Ruminococaceae, Helicobacteriaceae, and Pseudomonadaceae populations were higher in 5XFAD-Tg mice than in control mice. Prevotellaceae populations were lower in 5XFAD-Tg mice than in control mice. However, treatment with NK46 reduced Ruminococaceae, Lachnospiraceae, Helicobacteriaceae, and Pseudomonadaceae populations and increased the Prevotellaceae population. At the species level, EF603109_s, EU622763_s, and EU622668_s populations were lower in 5XFAD-Tg mice than in control mice while DQ815759_s, Helicobacter mesocricetorum group, and DQ815411_s populations were higher. However, treatment of 5XFAD mice with NK46 increased EF603109_s and EU622763_s populations and reduced DQ815411_s population. Furthermore, NK46 treatment increased β-diversities (Fig. 2C). To understand the balance of beneficial to harmful bacteria in the gut microbiota composition, we investigated the ratio of live Enterobacteriaceae to Bifidobaceria plus Lactobacilli (E/BL) by using selective media, glucose blood liver (BL) and deoxycholate hydrogen sulfide lactose (DHL) agar plates (Fig. 2D). The E/BL ratio was higher in 5XFAD-Tg mice than in control mice. However, NK46 treatment reduced the E/BL ratio. Furthermore, treatment with NK46 inhibited fecal bacterial LPS production in 5XFAD-Tg mice (Fig. 2E).
Next, we examined whether NK46 treatment could suppress gut inflammation in 5XFAD-Tg mice (Fig. 2F). Myeloperoxidase activity and NF-κB activation significantly increased in the colons of 5XFAD-Tg mice compared to those of the control mice. However, oral administration of NK46 suppressed myeloperoxidase activity and IL-6 and TNF-α expression in the colon of 5XFAD-Tg mice ( Fig. 2F(a-c)). NK46 treatment also inhibited NF-κB activation and COX-2 expression and increased the expression of claudin-1, a tight junction protein, in the colon of 5XFAD-Tg mice ( Fig. 2F(d)).
Next, we examined whether NK46 could suppress gut inflammation in aged mice. Myeloperoxidase activity and NF-κB activation significantly increased in the colons of aged mice more potently than in those of control mice (Fig. 4F). However, oral administration of NK46 suppressed myeloperoxidase activity and TNF-α and IL-6 expression in the colon of aged mice ( Fig. 4F(a-c)). NK46 treatment also inhibited NF-κB activation and COX-2 expression and increased claudin-1 expression in the colon of aged mice (Fig. 4F(d)).
NK46 attenuated cognitive decline in aged mice. To understand whether NK46 could alleviate aging-dependent cognitive decline, we examined the effect of NK46 in aged mice in the novel object recognition, Y-maze, and passive avoidance tasks (Fig. 5). The cognitive function of 19-month-old (aged) mice was significantly impaired compared with that of 7-month-old (adult) mice (Fig. 5A-C). Oral administration of NK46 for 1 month in 18-month-old mice significantly alleviated aging-dependent cognitive decline. NK46 treatment also suppressed aging-induced TNF-α and IL-6 expression in the hippocampus (Fig. 5D,E). Aging also significantly increased Iba + , LPS + /CD11b + , and caspase-3 + /NeuN + cell populations in hippocampus (Fig. 5F-H). However, treatment of aged mice with NK46 significantly suppressed the infiltration of Iba + , LPS + /CD11b + , and caspase-3 + /NeuN + cells into the hippocampus. Aging also suppressed BNDF and claudin-5 expression and CREB phosphorylation and induced NF-κB activation and caspase-3 and p16 expression in the hippocampus (Fig. 5I). However, NK46 treatment alleviated aging-dependent suppression of BDNF and claudin-5 expression and CREB phosphorylation and induction of NF-κB activation and p16 expression. Treatment of aged mice with NK46 also suppressed blood TNF-α and LPS levels (Fig. 5J,K).

Discussion
Aging is strongly associated with inflammation. Aging-dependent exposure to chronic, low-grade inflammation, termed "inflammaging", triggers aggressive neurodegenerative diseases such as AD 30,31 . The pathological hallmarks of AD are Aβ plaques and hyper-phosphorylated tau tangles, which are accelerated by microbial infection 30,32 . Moreover, aged people and mice exhibit altered gut microbiota composition compared to children/adults and young mice, respectively 17,33 . The Bifidobacteria population is decreased in the elderly 34 . Both Bifidobacteria and Lactobacilli populations are lower in elderly individuals than in adults, whereas there are no differences in Bacteroides and Eubacterium levels 35 . The gut microbiota composition of the elderly is extremely variable between individuals 36 . Elderly subjects exhibit high populations of Escherichia coli and Bacteroidetes 37 . The Firmicutes population and fecal LPS are significantly higher in aged mice than adults and Proteobacteria population and fecal LPS production are significantly higher in 5XFAD-Tg mice 19 . 5XFAD-Tg and aged mice exhibit altered gut microbiota composition, as well as increased proinflammatory cytokine expression in the GI tract 18,19 . These results suggest that the excessive production of gut bacterial LPS by gut dysbiosis may cause GI inflammation. Jang et al. reported that GI inflammation by 2, 4, 6-trinitrobenzenesulfonic acid deteriorated cognitive function with gut dysbiosis 29 . These results suggest that GI inflammation can increase memory impairment.
In the present study, we also found that 5XFAD-Tg and aged mice exhibited the Proteobacteria population more abundantly than adult control mice. Furthermore, the production of gut bacterial LPS were higher in the colon fluid and the expression of tight junction protein such as claudin was lower in the colon of 5XFAD-Tg and aged mice than in control mice. NF-κB activation and COX-2 and TNF-α expression were higher in the colon of 5XFAD-Tg and aged mice than in the colon of adult mice. We also found that blood LPS levels were higher in 5XFAD-Tg and aged mice than in control mice. Jang et al. reported that GI inflammation increased the absorption of gut microbiota LPS into the blood in mice and LPS treatment caused endotoxemia as well as colitis in mice 22 . In the present study, Tg mice exhibited excessive Aβ plaques and increased gut Proteobacteria and Firmicutes populations than aged mice, while aged mice induced the p16 expression and ration of Enterobacteriaceae population to Lactobacilli plus Bifidobacteria population more potently than 5XFAD-Tg mice. These results suggest that the gut microbiota of Tg and aged mice excessively induce LPS production and the chronic exposure to excessive gut microbiota LPS can cause endotoxemia through the acceleration of GI inflammation.
Moreover, we found that NF-κB activation and caspase-3 + /NeuN + (apoptotic neuron cell), Iba + (activated microglia), and LPS + /CD11b + (LPS-phagocytic cell) populations were increased in the hippocampus of 5XFAD-Tg and aged mice. These results suggest that aging and hippocampal Aβ plaque accumulation can cause hippocampal inflammation via the activation of microglia cells by gut microbiota endotoxins such as LPS. However, the expression of BDNF, which maintains synaptic plasticity and memory storage in the hippocampus 38,39 , and claudin-5, which is one of tight junction protein in the brain 40 , was suppressed in these mice. Brain inflammation causes the suppression of BDNF and caludin-5 expression 38,40 . These results suggest that hippocampal inflammation in 5XFAD-Tg and aged mice can be accelerated by the suppression of tight junction protein expression in the brain as well as GI tract. In the present study, we found that BDNF expression and cognitive function were suppressed in 5XFAD-Tg and aged mice compared to those in control mice. Devi and Ohno reported that the formation of Aβ plaques reduced BDNF levels in 5XFAD-Tg mice 41 . Rangasamy et al. also reported that the overexpression of Aβ proteins increased NF-κB activation in 5XFAD-Tg mice 42 . Gut microbiota LPS levels in the intestine fluid and NF-κB activation in the hippocampus are higher in 5XFAD-Tg mice than in control mice 19 . LPS levels in the colon fluid and blood and NF-κB activation in the hippocampus are also higher in aged mice than in young mice 18 . These results suggest that hippocampal inflammation can suppress BDNF expression, resulting in memory impairment through the suppression of NF-κB-mediated BDNF expression.
However, treatment with NK46, which suppressed gut microbiota LPS production and LPS-induced NF-κB activation in vitro, suppressed NF-κB activation, COX-2 expression, and myeloperoxidase activity in the colon of 5XFAD-Tg and aged mice and increased tight junction protein expression. NK46 treatment reduced gut microbiota LPS production and Proteobacteria population in 5XFAD-Tg and aged mice. These results suggest that NK46 www.nature.com/scientificreports www.nature.com/scientificreports/ can alleviate Aβ-and aging-induced GI inflammation through the regulation of gut microbiota composition and LPS production.
NK46 treatment also suppressed LPS levels in the blood and feces of 5XFAD-Tg and aged mice and alleviated cognitive decline in 5XFAD-Tg and aged mice. NK46 treatment suppressed NF-κB activation, Aβ, caspase-3, β-secretase, and γ-secretase expression, Aβ plaque accumulation, and the number of caspase-3 + /NeuN + (apoptotic neuron cells), Iba + (activated microglia), and LPS + /CD11b + (LPS-phagocytic) cells and increased BDNF Figure 5. NK46 attenuated cognitive decline in aged mice. Effects on exploration time in novel object recognition task (A), spontaneous alteration in Y-maze task (B), and latency time in passive avoidance task (C). Effect on IL-6 (D) and TNF-α (E) expression in hippocampus, assessed by ELISA. Effect on Iba1 + (F), LPS + / CD11b + (G), and caspase-3 + /Neu + cell populations (H) in hippocampus, assessed by a confocal microscope. Bar indicates 0.1 mm. (I) Effects on BDNF, caspase-3, and claudin-5 expression, NF-κB and CREB activation, assessed by immunoblotting. Effect on TNF-α (J) and LPS levels (K) in the blood. Test agent (Ag, vehicle alone; NK46, 1 × 10 9 CFU/mouse/day) was orally administered for 1 month in aged mice. Control adult mice (Con) were treated with vehicle alone. All data were expressed as mean ± SD (n = 6). # p < 0.05 vs. Con group. * p < 0.05 vs. Ag group. www.nature.com/scientificreports www.nature.com/scientificreports/ expression in the hippocampus of 5XFAD-Tg and aged mice. βand γ-Secretases and caspase-3 catalyze Aβ plaque formation, leading to neuron cell death 6,7 . LPS-induced neuroinflammation increases β-secretase and γ-secretase expression in mice 43 . Moreover, the inhibition of β-secretase and γ-secretase expression and induction of BDNF expression by Lactobacillus plantarum C29 treatment alleviates memory impairment in 5XFAD-Tg mice 19 . These results suggest that NK46 can improve cognitive function in 5XFAD-Tg and aged mice by increasing BDNF expression and suppressing NF-κB activation and β-secretase and γ-secretase expression, which could be both dependently and independently induced by the Aβ formation. NK46 significantly suppressed LPS-induced NF-κB activation in in vitro study. Furthermore, NK46 suppressed gut microbiota LPS-induced NF-κB activation in the colon and hippocampus. These results suggest that NK46 and its byproducts such as lipoteichoic acids and short chain fatty acids may inhibit gut dysbiosis and bacterial LPS production, resulting in the attenuation of the gut inflammation. Overall, NK46 alleviated gut dysbiosis, colitis, and cognitive decline in 5XFAD-Tg mice more potently in aged mice. These results may be due to short-term (4 weeks) treatment with NK46 in aged mice compared to its treatment for 8 weeks in 5XFAD mice.
Conclusively, our finding support the suggestion that gut dysbiosis and their excessive endotoxin production may cause endotoxemia and systemic inflammation, resulting in neuropsychiatric disorders: neuropsychiatric disorder-inducible stressors can alter gut dysbiosis through the regulation of HPA axis and gut dysbiosis can cause psychiatric disorders through the regulation of MGB axis [44][45][46] . Moreover, Bifidobacteirum longum can modify gut microbiota, particularly the Proteobacteria population, and their LPS production in 5XFAD-Tg and aged mice and suppress the progression of GI inflammation and endotoxemia, resulting in the attenuation of cognitive decline in 5XFAD-Tg and aged mice with the regulation of neuroinflammation via MGB axis.
To decide the dose of NK46 in mouse experiments, NK46 (1 × 10 8 and 1 × 10 9 CFU/mouse/day) were orally treated for 5 days in immobilization stress-treated mice and its anti-colitis effects such as colon length and macroscopic score were measured according to the method of Lee et al. 19 . Treatment with NK46 at a dose of 1 × 10 9 CFU/mouse/day alleviated more potently than at a dose of 1 × 10 8 CFU/mouse/day. Therefore, we orally gavaged NK46 at a dose of 1 × 10 9 CFU/mouse/day for the further in vivo study.

Animals.
All animal experiments were approved by the Committee for the Care and Use of Laboratory Animals in the Kyung Hee University (KHUASP(SE) 17-029 and 17-128) and were performed according to the NIH and Kyung Hee University Guidelines for Laboratory Animals Care and Use. Mice were fed a standard laboratory diet, allowed to take water ad libitum, and maintained in a ventilated room (temperature, 22 °C ± 1 °C; humidity, 50% ± 10% humidity); and a 12-h diurnal light cycle, 07:00-19:00) for 2 months before the animal experiment. They were housed in wire cages (3 mice/cage).
5XFAD-Tg mice (6 months-old) were separated into two groups, which were treated with vehicle or NK46 (1 × 10 9 CFU/day/mouse) six times per week for 8 weeks, as previously reported 47 . Control mice (6 months-old) were treated with vehicle. Each group consisted of six mice.
Aged mice (18 months-old) were also separated into two groups, which were treated with vehicle or NK46 (1 × 10 9 CFU/day/mouse) six times per week for 4 weeks. Control mice (6 months-old) were treated with vehicle for 4 weeks. Each group consisted of six mice.
For the assays of biochemical parameters, mice were then anesthetized 2 h after performing the final task. Blood, brain and colon were removed. Plasma was prepared by centrifuging blood. Colons were opened longitudinally and gently cleared of stool using phosphate-buffered saline (PBS). These tissues were used for immunostaining, immunoblotting, and ELISA. Immunofluorescence assay. Immunostaining analysis of brain slices was performed according to the method of Duncan and Miller 49 and Jang et al. 22 . Microglial cells were visualized by staining with anti-Iba1 antibody (1:200, Santa Cruz). Apoptotic neuron cells were stained with anti-caspase-3 and anti-NeuN antibodies (1:500, Millipore). LPS were stained with ant-LPS antibody (1:200, Abcam). Briefly, the brains were cryoprotected in 30% sucrose-PBS and then frozen with optimal cutting temperature compound and stored at −80 °C until processed. Brain tissue blocks were cryosectioned at a thickness of 30 μm, stored at 4 °C in the storing solution (30% ethylene glycol in PBS), permeabilized in 0.5% Triton X-100 for 5 min, blocked in 10% bovine serum with tween 20-contained PBS for 30 min, and incubated for 16 h at 4 °C with antibodies. Secondary antibodies conjugated with Alexa Fluor 488 (1:1,000, Invitrogen) or Alexa Fluor 594 (1:500, Abcam) were then treated to visualize. Nuclei were stained with DAPI. Immunostained samples were scanned with a confocal laser microscope.
Immunoblotting. Brain and colon tissues and cultured cells were homogenized with RIPA lysis buffer containing 1% protease inhibitor cocktail and a phosphatase inhibitor cocktail on the ice and centrifuged (13,200 g, 10 min, 4 °C) 19,48 . Proteins of supernatants were electrophoresed on sodium dodecyl sulfate-polyacrylamide gel, transferred to nitrocellulose membrane, blocked with non-fat dried-milk proteins, probed with antibodies for BDNF, CREB, p-CREB, p65, p-p65, COX-2, p16, caspase-3, Aβ, and β-actin, washed with PBS containing tween 20, and treated with secondary antibodies conjugated with horseradish peroxidase. Protein bands were visualized with an enhanced chemiluminescence detection kit.
Myeloperoxidase activity assay. Colon tissues were homogenized in 10 mM potassium phosphate buffer (pH 7.0) containing 0.5% hexadecyl trimethyl ammonium bromide, and centrifuged (13,200 g, 10 min, 4 °C). The resulting supernatants (50 μL) were added to the reaction mixture containing 0.1 mM H 2 O 2 and 1.6 mM tetramethyl benzidine preincubated at 37 °C for 2 min, and sequentially monitored the absorbance (650 nm) at 37 °C for 5 min 48 . Myeloperoxidase activity was calculated as the quantity of enzyme degrading 1 μmol/mL of peroxide, and expressed in unit/mg protein.

Determination of LPS.
Blood and fecal LPS levels were measured according to the methods of Kim et al. 50 .
For the assay of blood LPS contents, bloods collected by retroorbital bleeding into ethylenediaminetetraacetic acid-coated BD Microtainer ® tubes (Becton Dickinson, Franklin Lakes, NJ, USA) were centrifuged (13,200 g, 15 min). The supernatant (5 μL) was diluted 1:10 in pyrogen-free water and inactivated for 10 min at 70 °C. For the assay of fecal LPS contents, mouse feces were placed in 50 mL of PBS in a pyrogen-free tube, sonicated for 1 h on ice, and centrifuged (400 g, 10 min). The supernatant was collected, filtrated through a 0.45 μm Millipore filter, re-filtrated through a 0.22 μm filter, and inactivated at 70 °C for 10 min. Each filtrate or supernatant (50 μL) was incubated with LAL solution at 37 °C for 30 min, added additional reagents to formation of a magenta derivative, and measured the absorbance at 545 nm.
Culture of fecal bacteria. Fecal Enterobacteriaceae and Lactobacilli/Bifidobacteria populations were counted using the selective media, DHL and BL agar plates, according to the method of Kim et al. 50 .
Pyrosequencing. DNA was extracted from the fresh stools of mice (excluded mice trans-cardiacally perfused for brain tissue fixation) using a commercial DNA isolation kit (QIAamp DNA stool mini kit), as previously reported 20 . Genomic DNA was amplified using barcoded primers, which targeted the V3 to V4 region of the bacterial 16S rRNA gene. Pyrosequencing was carried out using a 454 GS FLX Titanium Sequencing System (Roche, Branford, CT) according to the method of Lee et al. 19 . Sequence reads were identified using the EzTaxon-e database (http://eztaxon-e.ezbiocloud.net/) on the basis of bacterial 16S rRNA sequence data. The number of sequences analyzed, observed diversity richness (operational taxonomic units, OTUs), estimated OTU richness (ACE and Chao1), and coverage were calculated using the Mothur program and defined considering a cut-off value of 97% similarity with the bacterial 16S rRNA gene sequences. Pyrosquencing reads have been deposited in the NCBI's short read archive under accession number SRX3921564~3921588. Statistical analysis. Experimental data are indicated as means ± standard deviation, and were statistically analyzed using one-way ANOVA followed by Duncan's multiple range test (P < 0.05).