Lactobacillus-derived extracellular vesicles counteract Aβ42-induced abnormal transcriptional changes through the upregulation of MeCP2 and Sirt1 and improve Aβ pathology in Tg-APP/PS1 mice

Mounting evidence suggests that probiotics are beneficial for treating Alzheimer’s disease (AD). However, the mechanisms by which specific probiotics modify AD pathophysiology are not clearly understood. In this study, we investigated whether Lactobacillus paracasei-derived extracellular vesicles (Lpc-EV) can directly act on neuronal cells to modify amyloid-beta (Aβ)-induced transcriptional changes and Aβ pathology in the brains of Tg-APP/PS1 mice. Lpc-EV treatment in HT22 neuronal cells counteracts Aβ-induced downregulation of Brain-derived neurotrophic factor (Bdnf), Neurotrophin 3 (Nt3), Nt4/5, and TrkB receptor, and reverses Aβ-induced altered expression of diverse nuclear factors, including the downregulation of Methyl-CpG binding protein 2 (Mecp2) and Sirtuin 1 (Sirt1). Systematic siRNA-mediated knockdown experiments indicate that the upregulation of Bdnf, Nt3, Nt4/5, and TrkB by Lpc-EV is mediated via multiple epigenetic factors whose activation converges on Mecp2 and Sirt1. In addition, Lpc-EV reverses Aβ-induced downregulation of the Aβ-degrading proteases Matrix metalloproteinase 2 (Mmp-2), Mmp-9, and Neprilysin (Nep), whose upregulation is also controlled by MeCP2 and Sirt1. Lpc-EV treatment restores the downregulated expression of Bdnf, Nt4/5, TrkB, Mmp-2, Mmp-9, and Nep; induces the upregulation of MeCP2 and Sirt1 in the hippocampus; alleviates Aβ accumulation and neuroinflammatory responses in the brain; and mitigates cognitive decline in Tg-APP/PS1 mice. These results suggest that Lpc-EV cargo contains a neuroactive component that upregulates the expression of neurotrophic factors and Aβ-degrading proteases (Mmp-2, Mmp-9, and Nep) through the upregulation of MeCP2 and Sirt1, and ameliorates Aβ pathology and cognitive deficits in Tg-APP/PS1 mice.


INTRODUCTION
Alzheimer's disease (AD) is a neurodegenerative disease that causes Aβ-induced neuropathology and cognitive deficits 1 .While there are genetic cases of AD, the vast majority of cases occur sporadically in aged individuals 2 .As the global elderly population grows, a proper strategy for AD treatment is urgently needed.A key mechanism in AD pathology is the accumulation of Aβ in the brain.Aβ accumulation is accelerated by an imbalance of Aβ production and clearance 3,4 .Aβ accumulation produces various neuropathological changes, including increased oxidative stress, neuroinflammatory responses, decreased neurogenesis, synaptic and neuritic deterioration, and neuronal loss 1,5,6 .Transgenic mice overexpressing mutant forms of the human β-amyloid precursor protein (APP) and presenilin (PS) genes exhibit Aβ accumulation, neuroinflammatory responses in the brain, and cognitive decline 6,7 , supporting the notion that Aβ-induced change is a critical mechanism in AD pathology.
Emerging evidence supports the role of the gut microbiota in regulating brain function and the pathogenesis of AD.Multiple pathways have been proposed to explain the effects of gut microbiota, including activation of resident immune cells in the gut epithelium to release cytokines, production of bacterial metabolites and signaling molecules, changes in intestinal membrane permeability, and propagation of bacterial amyloidseeding effects 8,9 .Recent studies, including our own, have demonstrated that extracellular vesicles (EVs) derived from Lactobacillus plantarum have neuroactive potential to reverse stress-induced downregulation of MeCP2, Sirt1, and neurotrophic factors in hippocampal neurons and to improve stress-induced depressive behavior in mice 10,11 .Lactobacillus paracasei-derived EVs exert anti-inflammatory effects against intestinal inflammatory responses in dextran sulfate sodium (DSS)-induced colitis 12 .However, it has not been tested whether Lactobacillus-derived EVs have bioactive potential to modify Aβ-induced pathology and improve behavioral deficits.
Histone modifications, such as acetylation and methylation at the N-terminal region of histones, can change chromatin structures, the results of which promote or suppress the expression of specific genes 13,14 .Methyl-CpG binding protein (MeCP2) binds to methylated CpG dinucleotides, which can regulate gene expression negatively or positively, in concert with histone modification factors or other nuclear factors 15,16 .Various epigenetic mechanisms including MeCP2 play roles in the expression of neurotrophic factors [17][18][19] .MeCP2 deletion or MeCP2 knockdown downregulates Bdnf expression in the hippocampus of mice [20][21][22] .Conversely, MeCP2 overexpression in cortical neurons or in the hippocampus of mice upregulates Bdnf expression 23,24 .Tg-APP/PS1 mice have reduced levels of neurotrophic factors in the hippocampus, which is associated with MeCP2 downregulation 22 .Sirt1 knockdown downregulates the expression of Bdnf, Nt3, and Nt4/5 in HT22 cells 10 .HDAC inhibitors and siRNA-mediated SUV39H1 knockdown upregulates BDNF expression 25,26 .BDNF levels are decreased in the hippocampus and cerebrospinal fluid (CSF) of patients with AD 27,28 .BDNF, NT3, and NT4/5 act through TrkB and TrkC 29 .Transgenic mice overexpressing Aβ have reduced expression levels of BDNF, whereas activation of TrkB receptors with BDNF mimetics improves cognitive decline in AD model mice 30,31 .These results suggest that proper targeting of epigenetic mechanisms could be a strategy to upregulate the expression of neurotrophic factors in the brain and to mitigate AD pathology in the brain.
In the present study, we investigated whether Lactobacillus paracasei-derived EVs can modify the expression of neurotrophic factors, and alleviate Aβ-induced pathology and cognitive decline in Tg-APP/PS1 mice.
Tg-APP/PS1 mice were maintained and handled in accordance with The Animal Care Guidelines of Ewha Womans University, and the experimental procedure for EV treatment in this study was approved by the Ewha Womans University Animal Care and Use Committee (IACUC 16-019).

Preparation of Lactobacillus paracasei-derived EVs
Bacterial culture and EV isolation were carried out as previously described 12 .In brief, Lactobacillus paracasei was cultured in MRS broth (MBCell, Seoul, Republic of Korea) for 18 h at 37 °C with gentle shaking (150 rpm).When the optical density of the culture reached 1.0 at 600 nm, the bacterial culture was centrifuged at 10,000 × g for 20 min, and the supernatant was collected and passed through a 0.22-μm bottle-top filter (Corning, NY, USA) to remove remaining cells or cell debris.The filtrate was concentrated using a MasterFlex pump system (Cole-Parmer, IL, USA) and a 100-kDa Pellicon 2 Cassette filter membrane (Merck Millipore, MA, USA) and was subsequently passed through a 0.22-μm bottle-top filter.Lactobacillus paracasei-derived EVs were obtained from the resulting filtrate by centrifugation at 150,000 × g for 3 h at 4 °C.Pellets were washed and resuspended in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 1.8 mM KH 2 PO 4 ).The protein concentration of the resuspended EV fractions was determined using a BCA protein assay kit (Thermo Fisher Scientific, MA, USA).Collected EVs were stored at −80 °C until use.EVs from Lactobacillus paracasei were named Lpc-EV.
To induce MeCP2 overexpression in HT22 cells, the pCNS-D2-MeCP2 plasmid DNA (0.125 μg) and Lipofectamine-2000 (10 μl) were separately diluted in 250 μl of Opti-MEM® Medium (31985070; Gibco-Thermo Fisher Scientific, Paisley, Scotland, UK) and incubated for 5 min at RT. Diluted plasmid DNA and Lipofectamine-2000 (250 μl each) were mixed at a 1:1 ratio and incubated for 20 min at room temperature.The plasmid DNA and Lipofectamine mix (500 μl) were dropped onto HT22 cells in each well of a 6-well plate containing 1.5 ml of DMEM (no FBS and no antibiotics).After 6 h, the medium was replaced with DMEM containing 1% FBS, and HT22 cells were grown further for 18 h.MeCP2 expression levels were determined using RT-PCR at the indicated time points.Aβ42 (1 μM, final) or Aβ42 plus Lpc-EV (10 μg per ml, final) were treated starting 24 h after transfection.The pCNS-D2-MeCP2 plasmid was obtained from Korea Human Gene Bank, Medical Genomics Research Center, KRIBB, Republic of Korea.
Purified total RNA (400 ng) with a 28 S/18 S rRNA ratio of 1.8~2.0 and an A260/280 ratio of 1.8~1.9 was converted to firstand second-strand cRNA, which was then converted to biotin-labeled cRNA samples.Purified biotinlabeled cRNAs (7.5 µg) were fragmented in an array fragmentation buffer by heating to 94 °C for 35 min.Each of the fragmented, biotin-labeled cRNA samples (6.0 µg) was hybridized to a GeneChip™ Mouse Gene 2.0 ST Array, representing 33,793 mouse gene transcripts.After washing, the array signals were amplified with Amersham Fluorolink streptavidin-Cy3 (GE Healthcare Bio-Sciences, Little Chalfont, UK) and scanned using GeneChip® HT Scanner and AGCC software (Affymetrix GeneChip® Command Console, Version 3.2.2).
Microarray signals were converted into log2 scale values and normalized using a robust multi-array average (RMA) method implemented in Affymetrix® Power Tools (APT).The false discovery rate (FDR) was controlled by adjusting the p value using the Benjamini-Hochberg algorithm.The log 2 values of microarray data were used to account for the differences in the expression levels.Macrogen Inc. (Seoul, South Korea) was requested to perform GeneChip hybridization and collect raw data, including signaling reading of the internal quality control probes and extraction of the scanned raw data.

Gene Ontology enrichment analysis
Normalized microarray signal values between two experimental groups were ranked by signed log 10 -transformed t-test p values and analyzed using the Rank-Rank Hypergeometric Overlap (RRHO), as previously described 36,37 .A RRHO heatmap of microarray signal values was constructed using the RRHO website (http://systems.crump.ucla.edu/rankrank/).The ranked value was colored red if it was higher than the value of the comparison group, and blue if it was lower.A total of 33,793 signal values were analyzed and plotted on a RRHO geographic map.Genes that were up-or downregulated by Aβ and their expression was reversed by Lpc-EV were selected using a top 20% cutoff.Identified genes were grouped into functional clusters using k-means clustering.Gene Ontology (GO) enrichment analysis was used to determine whether any clusters assigned by k-means clustering contained genes that were involved in key biological processes, such as nervous system development or neurogenesis, regulation of DNA-binding transcription factors, chromatin modification and/or histone modification.The GO term hierarchy was assigned based on the Mouse Genome-Database (http:// www.informatics.jax.org), and the interactions among the selected genes were assessed using the STRING database (http://string-db.org).

Administration of Lpc-EV to mice
Lpc-EV was administered to mice as previously described 10,11 .Lpc-EV was orally administered to mice at a dose of 2.27 mg/Kg/day by drinking water from 6.5 months of age until sacrifice at 8.0 months of age.A water bottle containing Lpc-EV diluted in drinking water to the concentration of 15 μg/mL (1.29 × 10 9 EV particles/mL) was presented to mice in a regular home cage.Lpc-EV-containing bottles were replaced every other day.

Quantitative real-time PCR
Quantitative real-time PCR (qPCR) was carried out as previously described 22,33 .Briefly, HT22 cells cultured in a 6-well plate were harvested using TRIzol reagent (15596-018, Invitrogen).Hippocampal tissues were homogenized in TRIzol solution using pellet pestles (Z359971, Sigma-Aldrich, Saint Louis, MO, USA), and total RNA was isolated from the homogenates.Two micrograms of total RNA were converted into cDNA using a reverse transcriptase system (Promega, Madison, WI, USA).qPCR reaction mixture contained 4 μl of 1/8 diluted cDNA, 10 μl of 2X iQTM SYBR Green Supermix (Bio-Rad Laboratories, Foster City, CA, USA), and 1 μl each of 5 pmol/μl forward and reverse primers in 20 μl.qPCR was carried out using the CFX 96 Real-Time PCR System Detector (Bio-Rad Laboratories).Transcript levels were normalized relative to Gapdh and L32 levels.

Thioflavin S staining of Aβ deposition
Thioflavin S (ThS) staining was carried out as previously described 22 .Briefly, ThS (T1892, Sigma-Aldrich) was dissolved in 50% ethanol followed by dilution in H 2 O to 1 mM.Free-floating brain sections were washed with 1X PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 1.8 mM KH 2 PO 4 ) and mounted onto a glass slide.The brain sections were incubated with 1 mM ThS for 5 min.Stained sections were washed in 100%, 95%, and 50% ethanol for 30 s each and then rinsed with 1X PBS twice.The sections were dried and cover-slipped with anti-fade fluorescent mounting medium (S3023, DAKO, Carpinteria, CA, USA).ThS-stained plaques were photographed using an Olympus BX51 microscope equipped with a DP71 camera.The number of plaques and the size of the stained area of the plaques were analyzed using MetaMorph Microscopy Automation & Image Analysis software (Molecular Devices, Sunnyvale, CA, USA).

Behavioral tests
Behavioral tests were carried out as previously described 22 .The behavioral activity of mice in each test was recorded using a video-tracking system (SMART; Panlab Harvard Bioscience, Holliston, MA, USA) or a webcam recording system (HD Webcam C210, Logitech, Newark, CA, USA).
Novel object recognition test.The novel objective recognition test (NOR) was carried out as previously described 22 .A subject mouse was presented with two identical objects (Object A; wooden blocks; 3.5 cm × 3.5 cm × 7 cm) in the open field (40 × 40 × 35 cm) and the time spent exploring each object was recorded for 10 min.This procedure was repeated twice and regarded as the familiarization phase.Two hours after the familiarization phase, one familiar object was replaced with a new object (Object B, a 100-ml glass flask containing fresh cage bedding at a 3-cm depth; 6 cm in diameter × 10 cm in height).The time spent exploring each object was recorded for 10 min.Fifteen minutes later, the familiar object (Object A) was moved to a new location near the wall to line up with the new object (Object B) relative to the wall.Then the subject mouse was placed in the open field, and the time spent exploring each object was recorded for 10 min.This test was regarded as the novel location recognition test (NLR).Twenty-four hours after the familiarization phase, the subject mouse was presented with the familiar object (Object A) and a third, new object (Object C, a plastic block made by stacking four 60-mm culture dishes with a black-tape band; 5.5 cm in diameter × 7.5 cm in height), and the time spent with each object was recorded for 10 min.

Water maze test
The water maze test was performed as previously described 22 .The water maze consisted of a circular pool (90 cm in diameter and 50 cm in depth) filled with water (24 °C) to a depth of 40 cm.The water was rendered opaque by the addition of nontoxic white paint.(Sargent ® White Art Tempera Paint).Spatial visual clues were provided with different symbols (circle, cross, oblique stripes, and checkered patterns, each with a diagonal of 40 cm) on each wall of the room.A circular platform (10 cm in diameter) was placed in the target quadrant, submerged approximately 1 cm below the surface of the water, and located in the middle of the center of the pool and the tank wall.
The acquisition phase consisted of 5 days of training trials; two trials with a 6-h interval per day.The training trial was initiated by placing the subject mouse at the start position in the pool by facing the mouse head to the wall of the tank, and the time and path taken to find the platform was recorded using a video-tracking system.A trial lasted until the mouse reached the platform or when 90 s had elapsed.If a mouse did not find the platform within 90 s, the mouse was placed on the platform for 15 s to help it acquire the context of the platform with respect to spatial cues.After the completion of each trial, the mouse was dried and returned to its home cage.This procedure was repeated 6 h later.Therefore, two trials were completed in a day.On Day 6, after the hidden platform was removed from the pool, the mouse was placed in the pool by facing the wall of the tank at the position opposite to the target quadrant and the time spent and path in the four quadrants were recorded over 60 s.Mice that did not move at a location for >30 s during the probe trial were excluded from the final data.After the probe trial test, mice were placed in the pool with the platform and a small visual flag (4.5 cm in height × 4.5 cm in width) in the target quadrant, and the time spent and path in the four quadrants were recorded over 60 sec to confirm the absence of visual or motor deficits.
Passive avoidance test.The passive avoidance test was carried out as previously described 22 .The test apparatus consisted of a lighted chamber (15 × 15 × 20 cm; 1500 lux) and a dark chamber (15 × 15 × 20 cm), each equipped with a metal grid floor (1 mm in diameter, 1 cm apart between grid).On the first day, a subject mouse was placed in the lighted chamber with the door opened and allowed to freely explore the equipment for 5 min.On Day 2, the mouse was placed in the light chamber.After 30 s, the middle door was opened, and the latency to enter the dark chamber was recorded, which was regarded as the pretest.When the mouse entered the dark room, the door was closed, and two successive electric foot-shocks (100 V, 0.3 mA of electrical shock delivered for 1.5 sec with a 3-second interval) were delivered through the grid floor.After shock-dark room conditioning, the mice were returned to their home cages.On Day 3 (24 h later), the mice were individually replaced in the lighted chamber, and the latency to entering the dark chamber was recorded, which was referred as the posttest.The posttest was repeated on Day 6 (72 h later) and Day 8 (120 h later).The total freezing time during the testing period on Day 3 was also analyzed.

Statistical analysis
Two-sample comparisons were performed using Student's t-test, and multiple comparisons were performed using one-way ANOVA followed by the Newman-Keuls post hoc test, two-way ANOVA or two-way repeatedmeasures ANOVA followed by the Bonferroni post hoc test.All data are presented as the mean ± SEM, and statistical significance was accepted at the 5% level.

RESULTS
Lactobacillus paracasei-derived EVs (Lpc-EV) counteracted Aβ42-induced downregulation of neurotrophic factors and TrkB in HT22 cells Recently we reported that Lactobacillus plantarum-derived EVs has the ability to upregulate the expression of neurotrophic factors and TrkB in HT22 neuronal cells.We investigated whether Lpc-EV produces similar effects in HT22 neuronal cells that have been treated with Aβ42.Treatment with Aβ42 in HT22 cells downregulated the expression of Bdnf, Nt3, Nt4/5, Ngf, and TrkB.In contrast, treatment with Lpc-EV blocked the Aβ42-induced downregulation of Bdnf, Nt3, Nt4/5, Ngf, and TrkB (Fig. 1a).
We used two independent approaches to identify genetic or epigenetic factors that regulate the transcriptional effects of Aβ42 and Lpc-EV on the neurotrophin system.First, we investigated whether any known nuclear factors mediate the reversing effects of Lpc-EV on the expression of neurotrophic factors.A number of genetic and epigenetic factors, including Creb1, Rest, and Pea3 [38][39][40] , and MeCP2 and Sirt1 19,20,22,26,33,38,41,42 have been studied for their role in regulating the expression of neurotrophic factors.Therefore, we selected a number of gene sets that included those characterized to regulate the expression of neurotrophic factors in previous studies, and included genes that are involved in the acetylation, deacetylation, methylation and demethylation of histones (ex, histone H3K9), which are important in regulating neuronal plasticity in various pathologies 10,19,26,33,38,40 .Aβ42 treatment in HT22 cells downregulated the expression of Mecp2, Creb1, and Pea3.In contrast, treatment with Lpc-EV blocked Aβ42-induced downregulation of Mecp2 and Creb1 but not Rest or Pea3 (Fig. 1b).We investigated whether Aβ42 and Lpc-EV treatment affect the expression of histone modification factors.Aβ42 downregulated the expression of the histone acetyltransferases p300 and CBP; the histone deacetylases (HDAC) Sirt1, Sirt5, and Sirt7; and the histone-lysine demethylase Kdm4a.Lpc-EV upregulated the expression of Hdac2 and the histonelysine methyltransferases Setdb1 and G9a.In contrast, Lpc-EV blocked the Aβ42-induced downregulation of Sirt1, Sirt5, and Kdm4a, and Aβ42-induced upregulation of Hdac2, Setdb1, and G9a (Fig. 1c-f).These results suggest that Lpc-EV has the ability to induce transcriptional responses of multiple epigenetic factors including Mecp2 and Sirt1.
Second, we investigated the effects of Lpc-EV on the transcriptional responses of genes induced by Aβ42 using a genome-wide approach.We used a microarray assay followed by Rank-Rank Hypergeometric Overlap (RRHO) 36,37 to identify differentially regulated genes after Aβ42 treatment and their changes were reversed by Lpc-EV.Of the 32,317 transcripts in the microarray, 11,003 transcripts (34%) were upregulated, and the altered expression of those transcripts was reversed by Lpc-EV treatment (Quadrant A).In addition, 9479 transcripts (29%) were downregulated by Aβ42 and their altered expression was reversed by Lpc-EV (Quadrant D) (Fig. 2a, b).The transcripts in Quadrants A and D (11,003 and 9479, respectively) were first annotated with Mus musculus genes, and then the top 20% of transcripts by rank based on expression differences were selected.This resulted in 1321 genes in Quadrant A and 1079 genes in Quadrant D. Finally, the transcripts with a high confidence interaction score of 0.7 or higher in the STRING database were selected.This resulted in 1197 genes in Quadrant A and 799 genes in Quadrant D (Fig. 2a, b).Serial K-means clustering and Gene Ontology (GO) enrichment analyses identified seven functional clusters of genes (Fig. 2c, d).The identified functional clusters had the following features (Fig. 2e, f; Supplementary Fig. 1).Cluster #6 contained genes covered by GO terms for "neurogenesis", which included Nt3, Ngf, Vegfc, and Wnt3a (Fig. 2e, g).Cluster #7 contained genes assigned by GO terms for "regulation of transcription, DNA-templated", "chromatin organization", and "histone modification", which Fig. 2 Genome-wide microarray analysis identified a list of genes that were differentially expressed by Aβ42 and their altered expression was reversed by Lpc-EV.a, b A hypothetical RROH map with up-or downregulated genes by Aβ42 or Lpc-EV (a).A hypergeometric map between the genes changed by Aβ42 (x-axis) and those by Lpc-EV (y-axis).The genes were ranked based on expression differences between two comparison groups by their log 10 -transformed t-test p values and plotted on the x-and y-axes (b).Quadrants A and D contained 11,003 and 9,479 transcripts, respectively.The top 20% of transcripts ranked on expression differences were selected and used for further analyses.Up and down arrows indicate upregulated and downregulated expression, respectively.c, d Serial K-means clustering of the selected genes and subsequent Gene Ontology (GO) enrichment analyses led to identify seven functional clusters (c), which were visualized by a combined PPI network (d).Clusters 6 and 7 contained functional groups of genes for neurotrophic factors (Cluster 6) and for transcription and epigenetic regulation (Cluster 7).e, h A summary of key features of Clusters 6 (e, g) and 7 (f, h).Top three functional groups assigned by GO terms in Cluster 6 (e) and the PPI of 75 genes assigned with a GO term for "neurogenesis" (g).Top three functional groups covered by GO terms in Cluster 7 (f) and the PPI of 98 genes assigned with a GO term for "regulation of transcription, DNA-templated" (h).
Cluster #1 contained genes involved in "Organic substance metabolic process", which included factors involved in RNA processing, and "Cellular component organization or biogenesis"; Cluster #2 contained genes involved in "Organic substance metabolic process", which included factors involved in DNA metabolic process, and "Cell cycle"; Cluster #3 contained genes involved in "Establishment of localization", which included factors involved in ion transport, and "Homeostatic process"; Cluster #4 contained genes involved in "Regulation of biological process", which included factors involved in regulation of signaling, and "Cell communication"; and Cluster #5 contained genes involved in "Cellular metabolic process", which included factors involved in carboxylic acid metabolic process, and "Nitrogen compound metabolic process" (Supplementary Fig. 1).These results suggest that Lpc-EV can block the changes in cellular and transcriptional responses that are induced by Aβ42.

Lpc-EV-induced upregulation of neurotrophic factors and TrkB in HT22 cells was mediated by epigenetic factors
Our initial investigation (Fig. 1) and genome-wide analysis (Fig. 2; Supplementary Fig. 1) showed that a number of genes with potentially diverse biological functions are involved in regulating Aβ42-induced changes and the effects of Lpc-EV.Of those identified factors, in this study, we focused on and investigated the changes in the expression of neurotrophic factors by Aβ42 and Lpc-EV and their potential upstream regulators.siRNAmediated Mecp2 knockdown in HT22 cells blocked Lpc-EV-induced upregulation of Bdnf, Nt3, Nt4/5, and TrkB.siRNA-mediated Sirt1 knockdown produced similar effects (Fig. 3a, b).siRNA-mediated Sirt5 knockdown blocked Lpc-EV-induced upregulation of Bdnf and TrkB but not Nt3 or Nt4/5 (Fig. 3c).siRNA-mediated Kdm4a knockdown weakly blocked the Lpc-EV-induced upregulated expression of Bdnf and Nt3 but not Nt4/5 or TrkB (Fig. 3d).
Overall, these results suggest that Lpc-EV induces the upregulation of Bdnf, Nt3, Nt4/5, and TrkB by transcriptionally activating or suppressing multiple epigenetic factors.Of these, MeCP2 and Sirt1 serve as common mediators for Lpc-EV-induced effects on the expression of these genes, whereas Sirt5, Kdm4a, Hdac2, and G9a are selectively involved in the regulation of a subset of the genes.
Overall, these results suggest that the Lpc-EV-induced upregulation of Bdnf, Nt3, Nt4/5, and TrkB is mediated by activation of multiple epigenetic factors that interact with each other, while partly converging onto Mecp2 and Sirt1.

Lpc-EV treatment restored the downregulated expression of neurotrophic factors in the hippocampus of Tg-APP/PS1 mice
We investigated whether Lpc-EV administration could change the expression of neurotrophic factors and TrkB in the brains of Tg-APP/PS1 mice.Compared to wild-type mice, Tg-APP/PS1 mice at 8 months of age had significantly downregulated expression of Bdnf, Nt4/5, and TrkB in the hippocampus.In contrast, Tg-APP/PS1 mice treated with Lpc-EV had upregulated expression of Bdnf, Nt4/ 5, and TrkB compared to that in wild-type mice (Fig. 4a, b).
Immunohistochemical analysis indicated that compared to wildtype mice, Tg-APP/PS1 mice exhibited downregulated expression of MeCP2 and Sirt1 in pyramidal and granule neurons of the hippocampus.In contrast, Tg-APP/PS1 mice treated with Lpc-EV showed upregulated expression of MeCP2 and Sirt1 in the hippocampus compared to that in the hippocampus of Tg-APP/ PS1 mice (Fig. 4e-i), suggesting that Lpc-EV treatment restored the downregulated expression of MeCP2 and Sirt1 proteins in the hippocampal neuronal neurons of Tg-APP/PS1 mice.

Lpc-EV treatment rescued the downregulated expression of
Mmp-2, Mmp-9, and Nep, and alleviated Aβ plaque accumulation in the brains of Tg-APP/PS1 mice Thioflavin-S (ThS) staining indicated that Lpc-EV treatment tended to reduce the number and total area of ThS-stained plaques in the parietal cortex and hippocampus of Tg-APP/PS1 mice compared to Tg-APP/PS1 mice (Fig. 5a-d).
We explored the cellular factors known to regulate Aβ production and Aβ clearance.The expression levels of Presenilin-1 (Psen1) and Psen2 in the hippocampus of Tg-APP/PS1 mice were slightly enhanced compared to those in wild-type mice, whereas the expression levels of Beta-secretase 1 (Bace1), Bace2, A Disintegrin and Metallopeptidase Domain 10 (Adam10) and Apolipoprotein E (Apoe) were not significantly changed.Tg-APP/PS1 mice treated with Lpc-EV had similar expression levels of those factors to Tg-APP/PS1 mice (Supplementary Fig. 4a).The expression levels of Matrix metalloproteinases-2 (Mmp-2), Mmp-9, tissue plasminogen activator (tPA), Insulin-degrading enzyme (Ide), Neprilysin (Nep), and Low-density lipoprotein receptor-related protein-1 (Lrp1) in the hippocampus of Tg-APP/PS1 mice were reduced compared to those in wild-type mice.In contrast, compared to Tg-APP/PS1 mice, Tg-APP/PS1 mice treated with Lpc-EV had significantly increased expression levels of Mmp-2, Mmp-9, and Nep but not tPA, Ide, or Lrp (Fig. 5e).
Lpc-EV treatment tended to suppress gliosis in the brains of Tg-APP/PS1 mice Immunohistochemical analysis showed that compared to wildtype mice, Tg-APP/PS1 mice at 8 months of age had increased Iba-1 staining in microglia throughout the brain.Lpc-EV treatment in Tg-APP/PS1 mice tended to reduce Iba-1 staining in microglia compared to that in Tg-APP/PS1 controls, but the reduction in the parietal cortex and hippocampus was not statistically significant (Fig. 6a-d).Quantitative analysis indicated that the levels of Iba-1stained microglia surrounding large plaques were not decreased following Lpc-EV treatment (Fig. 6e-g).In addition, GFAP staining indicated that Tg-APP/PS1 mice had increased astrogliosis throughout the brain, whereas Tg-APP/PS1 mice treated with Lpac-EV had significantly reduced levels of astrogliosis in the brain (Supplementary Fig. 5).

Lpc-EV treatment increased neurogenesis in the hippocampus of Tg-APP/PS1 mice
The relative MAP2 staining level in the dendritic processes of CA1 pyramidal neurons compared to that in the cell body region of CA1 pyramidal neurons in the stratum radiatum in Tg-APP/ PS1 mice was reduced compared to that in the stratum radiatum of wild-type mice.In contrast, compared to Tg-APP/PS1 mice, Tg-APP/PS1 mice treated with Lpc-EV had increased MAP2 staining levels in the dendritic processes of CA1 pyramidal neurons (Fig. 7a-c).
Tg-APP/PS1 mice at 8 months of age had fewer Ki-67 (a marker of proliferating cells)-positive cells and doublecortin (DCX; a marker of neuronal differentiation)-positive cells in the dentate gyrus compared than wild-type mice.In contrast, Tg-APP/PS1 mice treated with Lpc-EV had more Ki-67-positive cells and DCX-positive cells than Tg-APP/PS1 mice (Fig. 7d, e; Supplementary Fig. 6).
Lpc-EV treatment rescued cognitive deficits in Tg-APP/PS1 mice Next, we investigated whether Lpc-EV treatment improves cognitive deficits in Tg-APP/PS1 mice (Fig. 8a).In the novel object recognition (NOR) test (Fig. 8b), wild-type mice, Tg-APP/PS1 mice, and Tg-APP/PS1 mice treated with Lpc-EV explored two identical objects for a similar amount of time during the familiarization phase (Fig. 8c).In the NOR test two hours later (NOR-2 h), Tg-APP/ PS1 mice did not prefer the novel object over the familiar object, whereas Tg-APP/PS1 mice treated with Lpc-EV preferentially explored the novel object over the familiar object (Fig. 8d).In the novel location recognition test performed 15 min later (NLR-15 min), Tg-APP/PS1 mice did not prefer the displaced object, whereas Tg-APP/PS1 mice treated with Lpc-EV explored the displaced object over the novel object presented 15 min before (Fig. 8e).In the NOR test examined 24 h after familiarization (NOR-24 h), Tg-APP/PS1 mice did not preferably explore the novel object over the familiar object.In contrast, Tg-APP/PS1 mice treated with Lpc-EV preferentially explored the novel object over the familiar object (Fig. 8f).These results suggest that Tg-APP/PS1 mice treated with Lpc-EV show improved object recognition and retention memory.
Compared to wild-type mice, Tg-APP/PS1 mice exhibited an increase in latency to find the hidden platform compared to wildtype mice during the training phase of the water maze test.Tg-APP/PS1 mice treated with Lpc-EV showed a shortened latency to find the hidden platform compared to that of Tg-APP/PS1 control mice from Days 4 and 5 (Fig. 8g).In the following probe trial in which the escape platform was removed, Tg-APP/PS1 mice showed reduced exploration time in the target quadrant compared to that of wild-type mice, whereas Tg-APP/PS1 mice treated with Lpc-EV showed increased exploration time in the target quadrant compared to that of Tg-APP/PS1 mice (Fig. 8h, i).In the visual platform trial, wild-type mice, Tg-APP/PS1 mice, and Tg-APP/PS1 mice treated with Lpc-EV used similar amounts of times to reach the visual platform and similar swimming speeds during the test trial (Fig. 8j, k).
Compared to wild-type mice, Tg-APP/PS1 mice exhibited a reduced latency to enter the shock-associated dark chamber examined at 24 h, 72 h and 120 h after shock and reduced freezing time in the light chamber at 24 h after shock during the passive avoidance test.In contrast, Tg-APP/PS1 mice treated with Lpc-EV showed an increased latency to enter the shock-associated chamber until 72 h after shock and increased freezing time at 24 h after shock compared to that of Tg-APP/PS1 mice, suggesting that Lpc-EV treatment improved shock-associated retention memory (Fig. 8l, m).

Lpc-EV counteracted the Aβ-induced reduced expression levels of neurotrophic factors and TrkB through the upregulation of epigenetic factors
In the present study, we demonstrated that Lpc-EV treatment in HT22 cells upregulated the expression of Bdnf, Nt3, Nt4/5, and TrkB, along with upregulating or downregulating a number of nuclear or epigenetic factors, including Mecp2, Sirt1, and Hdac2 (Figs. 1, 2).Using systematic siRNA-mediated knockdown experiments, we provided evidence that Lpc-EV upregulated the expression of Bdnf, Nt3, Nt4/5, and TrkB through the transcriptional activation of those epigenetic factors (Fig. 3; Supplementary Figs. 2, 3).The results of the present study suggest the following interrelated issues regarding the roles of epigenetic factors in mediating Lpc-EV effects.
First, Lpc-EV upregulated the expression of Bdnf, Nt3, Nt4/5, and TrkB through activation of multiple epigenetic factors.Of the epigenetic factors, Mecp2 and Sirt1 were critical players in  Lpc-EV-induced upregulation of Bdnf, Nt3, Nt4/5, and TrkB (Fig. 3a, b), whereas Sirt5, Kdm4a, Hdac2, and G9a were partially or selectively involved in regulating the expression of Bdnf, Nt3, Nt4/5, and TrkB.The identified epigenetic factors were organized to form a hierarchical regulatory network, while partially converging onto Mecp2 and Sirt1 (Supplementary Fig. 7).Although the bioactive components of Lpc-EV remain unknown, it is possible that Lpc-EV cargo contains multiple types of bioactive components that activate several nodes in the regulatory network of the epigenetic factors (Supplementary Fig. 7).Considering that bacterial EVs contain various proteins, chemical metabolites, fatty acids, and nucleic acids [43][44][45] , candidate bioactive Lpc-EV cargo contents could be such bacterial components.Second, Lpc-EV counteracted Aβ42induced pathological changes.Lpc-EV treatment in HT22 cells changed the expression of Bdnf, Nt3, Nt4/5, TrkB, Mecp2, Sirt1, Sirt5, Kdm4a, Hdac2, and G9a in opposite directions to those induced by Aβ42 (Fig. 1a-f; Figs. 2, 3).It is worthwhile to understand the underlying mechanisms of how the two unrelated materials, Aβ42 and Lpc-EV, modulate the expression of the same epigenetic factors in opposite directions, but they remain to be elucidated.Third, our statistical analysis of Lpc-EV effects on HT22 cells indicates that Lpc-EV increased the transcript levels of Bdnf, Nt3, Nt4/5, Ngf, TrkB, Mecp2, Creb1, Sirt1, and Sirt5, and decreased the transcript levels of Setdb1, G9a, and Kdm4a (Fig. 1).These results raise the possibility that Lpc-EV can be used in the treatment of other brain disorders that are proceeded by reduced expression of neurotrophic factors, TrkB, Mecp2, and Sirt1.Lactobacillus plantarum-derived EVs (Lpl-EV) has the ability to counteract the stress-induced downregulation of neurotrophic factors, MeCP2, and Sirt1 in the brain and improves depressive-like behavioral deficits 10,11 .It remains to be determined whether Lpc-EV can exert protective effects against stress-induced depression-like behaviors, similar to those of Lpl-EV.
MeCP2-and Sirt1-dependent mechanisms in the regulation of neurotrophic factor expression have been reported in various cellular contexts.MeCP2 knockout mice have downregulated showing Iba-1-stained areas of microglia (red) surrounding ThS-stained plaques (green) in the parietal cortex of Tg-APP/PS1 mice (Tg) and Tg-APP/PS1 mice treated with Lpc-EV (Tg+Lpc-EV).Relative ratio of Iba-1-stained areas (f) and the intensity of Iba-1-stained microglia (g) over plaque areas.n = 8-9 animals.Iba-1-stained area ratio; t(15) = 0.7881, p = 0.4429; Iba-1-stained intensity ratio; t(15) = 1.677, p = 0.1143.Data are presented as mean ± SEM. *p < 0.05; **p < 0.01 (Student's t-test; and one-way ANOVA followed by the Newman-Keuls post hoc test).expression of Bdnf in the hippocampus 20,21 .Tg-APP/PS1 mice have reduced expression levels neurotrophic factors and MeCP in the hippocampus 22 .Consistently, MeCP2 binding to the promoter of the Bdnf, Nt3, and Nt4/5 genes is positively correlated with the expression levels of Bdnf, Nt3, and Nt4/5 in the hippocampus of Tg-APP/PS1 mice, and siRNA-mediated Mecp2 knockdown in the hippocampus downregulated the expression of Bdnf, Nt3, and Nt4/ 5 in the injected site 11 .Sirt1 regulates MeCP2 by deacetylation, which leads to an increase in Bdnf expression levels 46 .Sirt1 also upregulates Bdnf expression in the hippocampus in a CREBdependent manner 47 .In the present study, Mecp2 or Sirt1 knockdown blocked the Lpc-EV-induced upregulation of Bdnf, Nt3, Nt4/5, and TrkB in HT22 cells, whereas Mecp2 or Sirt1 knockdown produced subtle or no effect on Creb, respectively (Fig. 3f, h).Although the present study demonstrated the importance of MeCP2-and Sirt-1-dependent mechanisms in mediating Lpc-EV effects, it is possible that the epigenetic mechanisms regulating Lpc-EV effects on the brains of Tg-APP/ PS1 mice are more complex than those in HT22 cells.In addition, considering that the regulatory network composed of multiple epigenetic factors (Supplementary Fig. 7) might have a role in integrating other signaling pathways in neuronal and nonneuronal cells, we cannot rule out the possibility that MeCP2-or Sirt1independent pathways have a role in regulating Lpc-EV effects.Possible mechanisms underlying the Lpc-EV-induced alleviation of AD-like pathology in the brains of Tg-APP/PS1 mice treatment in Tg-APP/PS1 mice alleviated key AD-like pathology, including Aβ-plaque deposition (Fig. 5), neuroinflammatory responses (Fig. 6; Supplementary Fig. 5), and neurogenesis (Fig. 7; Supplementary Fig. 6), and improved cognitive deficits (Fig. 8).Although the detailed mechanism by which Lpc-EV produces such diverse effects in Tg-APP/PS1 mice needs to be further characterized, our results raise the following possibilities.
First, the Lpc-EV-induced upregulation of neurotrophic factors and TrkB might contribute to increased neurogenesis, neuroprotection of dendritic morphology, and enhanced cognitive function in Tg-APP/PS1 mice.BDNF regulates synaptic and behavioral plasticity [48][49][50][51] .However, the available evidence shows conflicting results.Viral vector-mediated BDNF expression in the entorhinal cortices improved hippocampal-dependent contextual fear conditioning, but it did not reduce Aβ accumulation in transgenic mice (J20 strain) that carry APP Indiana (V717F) and Swedish (K670M) mutations 52 .In contrast, the BDNF deficiency increased Aβ production and BDNF overexpression counteracted the downstream consequences of Aβ accumulation 53,54 .Therefore, it is worthwhile to study whether the Lpc-EV-induced upregulation of neurotrophic factors and TrkB modifies Aβ accumulation.
Third, the decreased neuroinflammatory responses induced by Lpc-EV might reduce Aβ pathology in the brains of Tg-APP/PS1 mice.Chronic neuroinflammation increases Aβ plaque deposition 56,57 , decreases hippocampal neurogenesis 58 , and induces cognitive decline 59,60 .Recently, we reported that Lpc-EV treatment in HT-29 human colorectal cancer cells downregulates the expression of LPS-induced pro-inflammatory cytokines IL-1α, IL-1β, IL-2, and TNFα; increases the expression of the antiinflammatory cytokines IL-10 and TGFβ; and attenuates intestinal inflammatory responses in dextran sulfate sodium (DSS)-induced colitis in C57BL/6 mice 12 .Therefore, it is possible that the Lpc-EVinduced suppression of neuroinflammatory responses contributes to improving pathological deficits in the brains of Tg-APP/PS1 mice, although direct evidence should be provided.
In conclusion, our results suggest that Lpc-EV has the ability to induce MeCP2-and Sirt1-dependent upregulation of Bdnf, Nt3, Nt4/5, TrkB, Mmp-2 and Mmp-9, and epigenetic modification is a critical mechanism by which Lpc-EV alleviates AD-like pathology in Tg-APP/PS1 mice.

Fig. 7
Fig. 7 Lpc-EV treatment increased neurogenesis and MAP-2-stained density of hippocampal dendritic processes in Tg-APP/PS1 mice.a-c Diagram of the regions examined for immunohistochemical analyses (a).Anti-MAP2 staining levels in the stratum radiatum of the indicated groups (b).Photomicrographs showing anti-MAP2-stained dendritic processes of pyramidal neurons in the stratum radiatum (c) in the CA1 region of WT-CON, Tg-CON and Tg-Lpc-EV mice.Higher magnification (low panels) of the boxed areas (c) of the indicated groups.The red box in the CA1 of (a) is the region for the images in (c) (upper panels).n = 9-12 animals.F(2,30) = 6.385, p = 0.0049.d, e The numbers of anti-doublecortin (DCX)-positive cells in the dentate gyrus (DG) of WT-CON, Tg-CON, and Tg-Lpc-EV mice (d).Photomicrographs showing anti-doublecortin (DCX)-stained cells in the dentate gyrus of the indicated groups (e).The red box in the DG of (a) is the region for the images in (e) (upper panels).Higher magnification (lower panels) of the boxed areas (e) of the indicated groups.n = 10-13 animals.F(2,31) = 5.136, p = 0.0118.Data are presented as the mean ± SEM. *, **, #, difference between the indicated groups.*, #, p < 0.05; **p < 0.01 (one-way ANOVA followed by Newman-Keuls post hoc test).