Accelerating action of Alzheimer’s disease gut microbiota on Tau protein hyperphosphorylation: crosstalk of inammation and autophagy

Background: The gut-brain axis has been implicated in the complex pathogenesis of Alzheimer’s disease (AD), but the action is unclear, this study was performed to clarify the effect of ad-related gut microbiota on the pathogenesis of AD during pregnancy and early exposure. Methods: A pilot study of gut microbiota in AD patients was performed. Gut microbiota structure, long-term potentiation (LTP), inammation levels, AD biomarkers, and metabolomics of serum and fecal were monitored after cohousing by cohousing in early life (from pregnancy 14 days to birth 14 days with 8-month old APP/PS1 mice).The regulatory action of bacterial metabolites on Tau protein phosphorylation was evaluated. Results: Gut microbiota from APP/SP1 mice altered structure of gut microbiota in newborn mice, LTP in hippocampal slices was signicantly shortened, and inammatory markers levels were increased, AD biomarkers were upregulated, with signicantly higher Tau protein phosphorylation at multiple sites (p < 0.05). Conclusions: These results imply ad-related gut microbiota can change the structure of gut microbiota during pregnancy and early exposure. Changing the structure of gut microbiota in the newborn mice can induce leucine metabolism disorder, induce mTOR mediated autophagy dysfunction and increase the level of inammation, thus leading to accelerate Tau protein phosphorylation and reduce LTP occurs in AD-cohousing mouse hippocampus.


Introduction
Gut microbiota interaction in diseases are becoming one of the most researched topics in life sciences.
Exponential growth data revealed correlations between dysbacteriosis and several neurological diseases including autism [1], depression [2], Parkinson's disease (PD) [3], and Alzheimer's disease (AD) [4]Therefore, researchers have gradually shifted their focus from targeting the central nervous system to targeting the characteristics of gut microbiota and/or intestinal homeostasis. cohousing until 14 days after birth, when the APP/PS1 mice were removed. The control group was treated the same as the model group, but APP/PS1 mice were replaced with 8-month-old C57 mice.

Electrophysiological recordings
Standard eld potential recordings were performed on the hippocampal cornu ammonis 1 (CA1) region using borosilicate glass micropipettes pulled to a tip diameter of about 1 µm and lled with 2 mol/L NaCl [18]. To record synaptic potentials, a recording electrode was placed at the CA1 apical dendrite region.
Stimulus intensity was set based upon input-output relationships and was 50% of the maximal response. For testing paired pulse facilitation (PPF), two stimuli with 50% of the maximal intensity were given at 15-, 50-, 100-, and 400-ms intervals. For recording long-term potentiation (LTP), stable baseline synaptic potentials (50% of the maximal intensity) were recorded for 20 minutes, and then a theta-burst tetanic stimulation that contained 15 burst trains at 5 Hz was delivered (each train contained ve pulses at 100 Hz). Thereafter, baseline intensity-evoked eld excitatory postsynaptic potentials (fEPSPs) were recorded for 60 minutes with 0.33 Hz. A custom bipolar platinum wire electrode (0.08-mm diameter) was placed at the Schaffer collateral pathway, and stimulation was delivered using a Model 2100 A-M Systems Isolated Pulse Stimulator (Carlsborg, WA, USA). All evoked responses were recorded using an Axoclamp-2B ampli er, and data acquisition was controlled with pClamp 10.2 software (Molecular Devices, Sunnyvale, CA, USA).

Pathological and physiological examination
After the mice sacri ced, the brain and intestinal tissues were dissected. A total of four brains from each group were xed in 4% paraformaldehyde solution and prepared as para n sections. Sections were stained with hematoxylin-eosin (H&E), and immuno uorescent assays were performed for ionized calcium-binding adaptor molecule 1 (IBA1) and glial brillary acidic protein (GFAP) using para nembedded 3 µm sections and a two-step peroxidase-conjugated polymer technique (DAKO Envision kit, DAKO, Carpinteria, CA). Slides were observed by light microscopy [4] .

Microbiome analysis
Fresh intestinal content samples were collected before the fasting of mice and stored at -80 °C. Frozen microbial DNA isolated from these samples had total masses ranging from 1.2 to 20.0 ng and were stored at -20 °C. The microbial 16S rRNA genes were ampli ed using the forward primer 5'-ACTCCTACGGGAGGCAGCA-3' and the reverse primer 5'-GGACTA CHVGGGTWTCTAAT-3', and the forward primer 5'-CCTAYGGGRBGCASCAG-3' and reverse primer 5'-GGACTACNNGGGTATCTAAT-3' for mice. Each ampli ed product was concentrated via solid-phase reversible immobilization and quanti ed by electrophoresis using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). After quanti cation of DNA concentration by NanoDrop, each sample was diluted to a concentration of 1 × 10 9 molecules/µL in Tris-EDTA buffer solution and pooled. Next, 20 µL of the pooled mixture was used for sequencing with Illumina MiSeq sequencing system according to the manufacturer's instructions (Illumina, San Diego, CA, USA). The resulting reads were analyzed as described previously [19].

Mice feces
A 40 mg feces sample was homogenized in 400 µL deionized water containing 10 µg/mL of L-norvaline as internal standard. Following centrifugation at 14,000 g and 4 °C for 15 min, a total of 300 µL supernatant was transferred. The extraction was repeated by adding 600 µL of ice-cold methanol to the residue. The supernatants from the two extractions were combined. A 400 µL volume of combined supernatants and 10 µL of internal standard solution (50 µg/mL of L-norleucine) were combined and evaporated to dryness under nitrogen stream. The residue was reconstituted in 30 µL of 20 mg/mL methoxyamine hydrochloride in pyridine, and the resulting mixture was incubated at 37 °C for 90 min. A 30 µL volume of Bis(trimethylsilyl)-tri uoroacetamide (BSTFA) (with 1% trimethylchlorosilane (TMCS)) was added into the mixture and derivatized at 70 °C for 60 min prior to gas chromatography-mass spectrometry (GC-MS) metabolomics analysis.
Metabolomics instrumental analysis was performed on an Agilent 7890A gas chromatography system coupled to an Agilent 5975C inert MSD system. An OPTIMA® 5 MS Accent fused-silica capillary column (30 m × 0.25 mm × 0.25 µm; MACHEREY-NAGEL, Düren, Germany) was utilized to separate the derivatives. Helium (> 99.999%) was used as a carrier gas at a constant ow rate of 1 mL/min through the column. Injection volume was 1 µL in split mode (2:1), and the solvent delay time was 6 min. The initial oven temperature was held at 70 °C for 2 min, ramped to 160 °C at a rate of 6 °C/min, to 240 °C at a rate of 10 °C/min, to 300 °C at a rate of 20 °C/min, and nally held at 300 °C for 6 min. The temperatures of the injector, transfer line, and electron impact ion source were set to 250 °C, 260 °C, and 230 °C, respectively. The electron ionization energy was 70 eV, and data were collected in a full scan mode (m/z 50-600).
The typical total ion current chromatograms are illustrated in Figure S1. The peak picking, alignment, deconvolution, and further processing of raw GC-MS data were from previously published protocols [20].
The nal data were exported as a peak table le, including observations (sample name), variables (rt_mz), and peak areas. The data were normalized against total peak abundances before performing univariate and multivariate statistics.
For multivariate statistical analysis, the normalized data were imported to SIMCA software (version 14.1, Umetrics, Umeå, Sweden), where the data were preprocessed with unit variance scaling and mean centering before performing principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and orthogonal PLDS-DA (OPLS-DA). The model quality is described by the R 2 X or R 2 Y and Q 2 values. R 2 X (PCA) or R 2 Y (PLS-DA and OPLS-DA) is de ned as the proportion of data variance explained by the models and indicates the goodness of t. Q 2 is de ned as the proportion of variance in the data predictable by the model and indicates the predictability of current model, calculated by a crossvalidation procedure. To avoid model over-tting, we performed a default 7-round cross-validation in SIMCA software to determine the optimal number of principal components.
For univariate statistical analysis, the normalized data were analyzed on the R platform (version 3.3.0). Parametric testing was performed on normally distributed data by Welch's t test, while nonparametric Wilcoxon Mann-Whitney tests were conducted on the data with abnormal distributions.
The variables with VIP values in the OPLS-DA model > 1 and p values for univariate statistical analysis < 0.05 were identi ed as potential differential metabolites (Fig. 8B). Fold change was calculated as a binary logarithm of the average normalized peak intensity ratio between Groups 1 and 2, where a positive value means that the average mass response of Group 1was higher than Group 2.

Mice serum
Serum was collected after the mice were sacri ced, and 80 µL serum was added into 240 µL of cold methanol with acetonitrile (2:1, v/v). Next, 10 µL internal tagging standard (L-2-chlorine-phenylalanine, 0.3 mg/mL, dissolved in methanol) was added, vortexed after 2 min, then extracted using ultrasonic extraction method for 5 min. After 20 min standing at -20 °C and centrifuging for 10 min (14000 prm, 4 °C), 200 µL of supernatant was loaded into a sample bottle with lining tube for liquid chromatographymass spectrometry (LC-MS) analysis (Waters UPLC I-class system equipped with a binary solvent delivery manager and a sample manager, coupled with a Waters VION IMS Q-TOF Mass Spectrometer equipped with an electrospray interface (Waters Corporation, Milford, MA, USA). LC Conditions: Column: Acquity BEH C18 column (100 mm × 2.1 mm i.d., 1.7 µm; Waters). Data analysis was performed as described previously for feces experiments.

RNA sequencing
We prepared RNA sequencing libraries using global brain samples and performed 150-bp paired-end sequencing using the Illumina HiSeq platform. RNA sequencing libraries were prepared from 2 µg of total RNA using the TruSeq Kit (Illumina) with the following modi cation: instead of purifying poly-A RNA using poly-dT primer beads, we removed ribosomal RNA using the Ribo-Zero rRNA Removal Kit (Illumina). All other steps were performed according to the manufacturer's protocols. RNAseq libraries were analyzed for quality control, and the average size of inserts was approximately 200 to 300 bp. The sequencing library was then sequenced on a Hiseq platform (Illumina).

Western blotting analysis
Cells were seeded into 6-well culture plates at 5 × 10 6 cells/well and washed twice with D-Hanks solution when the cells reached 80% con uence. The cells were harvested and lysed with protein lysis buffer, and protein concentrations were determined using a Coomassie Brilliant Blue G250 assay kit (Nanjing JianCheng Bioengineering Institute, Nanjing, China).

Statistical analysis
All data are means ± standard deviations (SD) of at least three independent experiments. Signi cant differences between treatments were analyzed by one-way analysis of variance at p < 0.05 using Statistical Package for the Social Sciences (SPSS, Chicago, IL, USA) and Prism 5 (GraphPad, San Diego, CA, USA) software.

Variation characteristics of gut microbiota in AD patients
Following ethics committee approval, feces was collected from 47 AD patients, 28 disease controls, and 26 healthy controls. The 16S rRNA analysis showed signi cant differences for the abundance-based covered estimator (ACE) and Chao 1 indexes (p < 0.05, Fig. 1A) from the healthy groups. The heatmap of genus abundance and the relative abundance of changed bacteria of the AD group show that the genera, Faecalibacterium, Fusicatenibacter, Roseburia, Lachnospira, Agathobacter, Megamonas and Prevotella 9 were decreased, while Klebsiella, Akkermansia, Rhodococcus, Pseudomonas and Escherichia-Shigella were increase compared to the healthy group (Fig. 1B). The three-phase diagram (Fig. 1C) showed that the abundances of some of bacteria species varied, with higher levels of conditioned pathogens in the disease groups, suggesting these bacteria may indicate an AD-related gut microbiota. To evaluate the role of gut microbiota on AD pathogenesis, we performed cohousing in mice.
The gut microbiota composition of normal C57 mice and APP/PS1 mice Before couhousing, we analyzed the gut microbiota composition of APP/PS1 mice using 16S rRNA sequencing. As shown in Fig. 2, the heatmap at genus ( Fig. 2A) levels could successfully distinguish the normal 8-month C57 and 8-month APP/PS1 mice. The relative abundances of Alistipes, Rikenella. Anaerotruncus, Turicibacter, Mucispirillum, Adlercreutzia, Desulfovibrio, Lactobacillus were increased compared to control, while Ruminococcus, Oscillospira and Clostridium, Prevotella, Sutterella, Anaerostipes and Bacillus were decreased ( Fig. 2A, p < 0.05). The results revealed that the gut microbiota composition of APP/PS1 mice were very different from the normal C57 mice.

Colonization of gut microbiota analysis
To explore the role of gut microbiota on AD pathogenesis, an cohousing model was used. After consulting the literature and considered our preliminary results, we used the methods of cohousing. after 2 weeks of pregnancy, the dams were placed with two 8-month-old APP/PS1. The mice were cohoused until 14 days after birth.
At rst and second time points, the gut microbiota composition of intestinal contents were analyzed. For the offspring of mice fed a standard diet, as shown in Fig. 3A, the operational taxonomic units (OTUs) of AD-cohousing mice (287 of S1N, 328 of S2N) were lower than that of control groups (318 of Control 1, 357 of Control 2). The heatmap of the relative abundance of genus showed obvious variation after cohousing with APP/PS1 mice (Fig. 3B). The ACE, Chao 1 and Shannon indexes showed the same trends, while groups S1N and S2N were lower than their corresponding control groups (S1N vs Control 1, S2N vs Control 2). Collectively, these results indicate that some bacteria could be colonized from APP/PS1 mice to newborn C57 mice during cohousing.
The dominant bacterial communities were the Firmicutes and Bacteroidetes ( Figure S1). The main differences in bacteria abundance at genus levels were decreases in uncultured bacterium Bacteroidales S24-7 group, Desulfovibrio, Turicibacter, Candidatus Saccharimonas, while Lactobacillus, Ruminiclostridium 5, Odoribacter, uncultured bacterium Porphyromonadaceae were increase (Fig. 3B, S1N vs Control 1, p < 0.05). The main differences in bacteria abundance at genus levels were decreases in Enterorhabdus, Ruminiclostridium, Desulfovibrio, Alloprevotella, uncultured bacterium Ruminococcaceae, while uncultured bacterium Mollicutes RF9, Lactobacillus, Candidatus Saccharimonas, Turicibacter and Ruminococcaceae UCG-014 were increased (Fig. 3B, S2N vs Control 2, p < 0.05).Those suggests that cohousing with APP/PS1 mice expands the differentiation of gut microbiota, or the bacteria from APP/PS1 mice could affect the microbe colonization that the normal C57 mice received from their mothers. A PCoA plot also show the different between S1N group and control1 group and between S2N group and control2 group( Figure S2). Lactobacillus and Turicibacter was increased both in cohousing mice and 8-month APP/SP1 mice, which may be the dominant species of colonization.

LTP (Long-term Potentiation) in hippocampal slices
Based on the colocalization of gut microbiota analysis using 16S rRNA, the mice were kept on their diets for 10 months, based on the fact APP/PS1 mice begin to exhibit learning and memory dysfunction at 6 months old [21]. Ten months later, subsets of mice were randomly selected to perform standard eld potential recordings. Repetitive stimulation (0.33 Hz) of Schaffer collaterals evoked fEPSPs in the hippocampal CA1 region (Figs. 4A, p < 0.05). The input-output relationship curves (Figs. 4B, p < 0.05) and linear slopes (4C, p < 0.05) revealed that compared to control mice (NC), the hippocampal CA1 neurons of cohousing mice showed decreased excitability. Synaptic short-term plasticity was measured using a PPF protocol in hippocampal slices. Statistical analyses from all slices demonstrated a signi cant inhibition of the ratio of P2/P1 at all tested P1 and P2 intervals in AD-cohousing mice compared to control mice of NC (Figs. 4D, p < 0.05), suggesting decreased hippocampal synaptic short-term plasticity. Finally, hippocampal CA1 synaptic LTP between AD-cohousing and control mice were compared. The plot recording time to normalized fEPSP slopes (baseline as 1) from pooled data showed impaired LTP induction (after theta-burst stimulation 0-10 min) and maintenance (after theta-burst stimulation 50-60 min) (Figs. 4E, p < 0.05, unpaired t test). The mean LTP induction results are shown in Figs. 4F and Figs. 4G (p < 0.05, unpaired t test).All these results suggested that reduced LTP occurs in AD-cohousing mouse hippocampus.

AD-cohousing enhances in ammation in brain and serum
Based on the gut microbiota structure changes and altered LTP in hippocampal slices after cohousing, we assessed brain morphology in mice. The remaining mice were sacri ced, and the brain sections were processed for H&E and Nissl staining, which showed obvious pathologic changes as the neuron shrinkage, neuron size and number reductions, and cytoplasmic vacuolar changes in AD-cohousing mice. Immuno uorescent labeling of microglial IBA-1 and GFAP in the CA1 area showed strong astrocyte and microglial activation (Fig. 5A).
To clarify the action of cohousing on AD pathogenesis, we rst assessed the AD biomarkers of Tau, Aβ4, β-site APP cleaving enzyme (BACE), and apolipoprotein E (APOE) in the brain. Their expression levels shifted after cohousing. Tau, Aβ4, BACE, and APOE were upregulated in most AD-cohousing mice(p < 0.05, Fig. 7A and Fig. 8A).
Abnormal hyperphosphorylation of Tau is positively correlated with cognitive dysfunction, neuro brillary degeneration and the degree of dementia [22]. we further detected the different phosphorylation sites of Tau protein using speci c antibodies. The difference was signi cant for p-Tau(Ser404) and p-Tau(Ser416) (p < 0.05, Fig. 7A), which indicated enhanced Tau protein hyperphosphorylation.These results indicate that AD gut microbiota could accelerate Tau protein hyperphosphorylation on multiple sites.

Fecal metabolites induce in ammation and autophagy in BV2 cells
To investigate the factors related to Tau protein hyperphosphorylation, we co-cultured BV2 cells with ADcohousing mice feces extracts. Then we performed western blotting to measure the oxidative stress biomarker glycogen synthase kinase 3 (GSK3)β, the in ammatory factor of nuclear factor (NF)-κB, the autophagy biomarkers of microtubule-associated protein light chain (LC)3A/B, and levels of Tau and p-Tau using western blotting. The results (Fig. 8A) showed that autophagy, in ammation, and oxidative stress were in uenced by the extracts of AD-like mice feces, which indicated that AD fecal metabolites can induce an imbalance of autophagy and activate in ammation and oxidative stress in BV2 cells.
PKA, PGLYRP2 and NOX1 (Fig. 8A) levels were altered by extracts of AD-cohousing mice feces (17.25 mg/ml high dose, AD-cohousing-H; 13.5 mg/ml of middle dose, AD-cohousing-M; and 7.7 mg/ml low dose, AD-cohousing-L; 13.5 mg/ml of middle dose, Control), which indicated GSK3β/protein kinase A (PKA) activation, which can lead to an imbalance in reactive oxygen species (ROS) and then aggravate oxidative damage and/or aging.

Metabolomics analysis of intestinal contents and serum in AD-cohousing mice
Based on the effects of extracts of AD-cohousing mice feces on BV2 cells, we performed metabolomics analysis of intestinal contents to investigate the target bacteria and/or metabolites. The metabolomics of intestinal contents were analysis using GC-MS, and serum was assessed with LC-MS. As shown in Fig. 8 and Fig. 9 .
Previous study have demonstrated that the leucine can activate the mTOR to regulate the autophagy and mitochondrial function [23]. In this study, we also found that the levels of leucine in feces were reduced, while norleucine levels were increased in serum, expressions of mTOR, TORC1 (CRTC1) and TORC2 (CRTC1) were up-regulated (Fig. 7A, p < 0.05). The difference of metabolism of feces and serum suggested that AD-cohousing disorder the leucine metabolism, make the mTOR unregulating the autophagy, then aggravates accumulation of pathological products and in ammation levels, but these still need more studies and clinical survey to con rm. Metabolites pathway analysis of differences metabolites of feces and serum between AD-cohousing mice and control in table 2 and table 3.

Discussion
Oxidative stress, in ammation, neurogenesis, immune responses, dysbacteriosis and infection all about above 30 factors have been associated with AD [24]. Gut microbiota can release signi cant amounts of harmful metabolites as amyloids and lipopolysaccharides, which might modulate signaling pathways and induce the production of proin ammatory cytokines related to AD pathogenesis. The gut microbiome, leaky gut, and bacterial translocation could be involved in AD. In this study, we also found evidence of dysbacteriosis in both AD patients and APP/PS1 mice (Figs. 1-3), especially an increase in pathogenic and conditioned pathogens in AD patients (Fig. 1). The behavior, oxidative stress, in ammation, neurogenesis, and immune response of AD-cohousing mice were altered compared to normal control mice (Figs. 3-7). However, some of them are bene cial or exert different functions dependent on the numbers of other bacteria, as members of the Alistipes genera show high abundance in the most frail individuals and in middle-aged and older mice and are decreased in autism spectrum disorders, colitis and colorectal cancer, chronic hepatitis B patients, and appendicitis [1,25]. Several studies reported that Bacteroides fragilis (B. fragilis), members of Bacteroides, and its immunomodulatory capsular polysaccharide A are equally effective in preventing colitis and experimental allergic encephalomyelitis in murine models [26]. These species also orchestrate robust protective anti-in ammatory responses during viral infections [22]. Rodents with high abundance of Alloprevotella genera in early life have high risk of behavioral phenotypes, with males but not females exhibiting de cits in social behavior [27]. Clinical investigation showed that the genera Lactobacillus, Clostridium IV, Paraprevotella, Clostridium sensu stricto, Desulfovibrio, and Alloprevotella were enriched in fecal samples from patients with chronic kidney disease [28].
Serum levels of many cytokines (TNF-α, EGF, IL-17α, prolactin, FGF-basic, MCP-1, IL-6, G-CSF, and MIP-1α) measured after cohousing (Fig. 5), and brain levels of in ammatory factors were also altered, including TREM2, IL-12α, IL-23, NF-κB p65, NLRP3 and TLP4 ( Fig. 6 and Fig. 7). Extracts of AD-cohousing mice feces in uenced levels of GSK3β, NF-κB, and LC3A/B in BV2 cells. Collectively, the results shows that gut microbiota change modulate signaling pathways and the production of proin ammatory cytokines related to AD pathogenesis. Although these limited microbiota animal models do not fully represent the situation in humans, there is considerable evidence of a role of gut microbiota in AD progression.
Recent human studies have investigated gut bacterial taxa and shown altered abundance in patients with AD [29]. A Chinese cohort had distinct microbial communities of Gammaproteobacteria, Enterobacteriales, and Enterobacteriaceae [4]. Li et al. studied AD patients and found that the fecal abundance of six genera increased (Dorea, Lactobacillus, Streptococcus, Bi dobacterium, Blautia, and Escherichia), while ve decreased (Alistipes, Bacteroides, Parabacteroides, Sutterella, and Paraprevotella) [28]. They also found that blood samples had different abundances between control and AD groups; Propionibacterium, Pseudomonas, Glutamicibacter, Escherichia, and Acidovora increased, while Acinetobacter, Aliihoe ea, Halomonas, Pannonibacter, Leucobacter, and Ochrobactrum decreased [28]. We found that the genera of Megamonas, Faecalibacterium,Lachnospira, Fusicatenibacter Prevotella 9 and Ruminococcus_2 were decreased, while Rhodococcus were increase compared to the control group (Fig. 1B). The commonly changed bacteria in AD-like mice and AD patients only belonged to the Lactobacillus genera, the more pathogenic bacteria in the AD patients were not found in speci c pathogen-free (SPF) AD-like mice, and few special bacteria found were different from previous studies [4,28,29]. L. plantarum-derived lactic acid triggered the activation of the intestinal NADPH oxidase Nox and ROS generation. In turn, ROS production promoted intestinal damage, increased intestinal stem cell proliferation, and dysplasia. Nox-mediated ROS production required lactate oxidation by the host intestinal lactate dehydrogenase, revealing host-commensal metabolic crosstalk that is probably broadly conserved [3]. The increase in intestinal permeability coincided with higher plasma levels of LPS, serum IL-1, and TNF-α. Clinical studies suggested that individuals with microbial dysbiosis due to intestinal diseases have a higher risk of AD [28]. In this study, the fecal metabolites of AD-cohousing mice in uenced PGRP-L and NOX1 expression in BV2 cells (Fig. 8A). Systemic in ammation was also observed. The study reveals marked sex differences in a multifactorial model of early-life adversity, both on emotional behaviors and gut microbiota and Lactobacillus genera was regulated by early adversity both in male and female [27]. Previous studies suggest that gut microbiota is associated with neuropsychiatric disorders, such as Parkinson'disease, amyotrophic lateral sclerosis, and depression. In the present study, clinical fecal samples were collected and analysed from AD patients for 16S ,which show gut microbiota is altered in AD patients and may be involved in the pathogenesis of AD [29].
Major group of microbes linked to AD include bacteria: Chlamydia pneumoniae, Helicobacter pylori, Porphyromonas gingivalis, Fusobacterium nucleatum, Prevotella intermedia, Actinomyces naeslundii, and spirochete group; fungi: Candida sp., Cryptococcus sp., Saccharomyces sp., Malassezia sp., Botrytis sp., and viruses: herpes simplex virus type 1 (HSV-1), human cytomegalovirus, and hepatitis C virus [10]. We also found that the pathogenic bacteria including Escherichia, Klebsiella, Pseudomonas, Rhodococcus, and Akkermansia were increased in AD patients, which indicated that these infections might an inducers and/or accelerators for oxidative stress, in ammation, autophagy, and neurodegeneration. Escherichia was increased at the genus level in both fecal and blood samples from subjects with AD and MCI [28]. Postmortem brain tissue from patients with AD showed that both LPS and gram-negative Escherichia coli fragments colocalize with amyloid plaques, and an increase in the abundance of a pro-in ammatory gut microbiota of Escherichia-Shigella and a reduction in anti-in ammatory Eubacterium rectale are possibly associated with a peripheral in ammatory state in patients with cognitive impairment and brain amyloidosis [10,28,30]. In this study, we found a higher abundance of Escherichia -Shigella in AD patients, and the expressions of GSK3β, NF-κB, LC3A/B, Tau, and p-Tau correlated with metabolites of feces of AD-cohousing mice in BV2 cells. IgG antibody levels to seven oral bacteria associated with periodontitis showed that abundance of Fusobacterium nucleatum and Prevotella intermedia were signi cantly increased at baseline serum draw in the AD patients compared to controls, and these remained signi cant when controlling for baseline age, Mini-Mental State Exam score, and APOE ε4 status, which suggested that chronic in ammation in periodontal disease could be a risk factor for AD [10,28,30]. F. nucleatum is an anaerobic oral commensal and a periodontal pathogen associated with a wide spectrum of human diseases. It is implicated in adverse pregnancy outcomes (chorioamnionitis, preterm birth, stillbirth, neonatal sepsis, preeclampsia), gastrointestinal disorders (colorectal cancer, in ammatory bowel disease, appendicitis), cardiovascular disease, rheumatoid arthritis, respiratory tract infections, Lemierre's syndrome, and AD [10,28,30]. Subtractive genomics analysis demonstrated that F. nucleatum infection could simultaneously regulate multiple signaling cascades that could upregulate proin ammatory responses, oncogenes, modulation of host immune defense mechanisms, and suppression of DNA repair system [28]. Microcin E492, a peptide naturally produced by Klebsiella pneumonia, assembles into amyloid-like brils in vitro, and these have the same structural, morphological, tinctorial, and biochemical properties as the aggregates observed in AD [10,28,30]. Low levels of the amino acid L-arginine in astrocytes surrounding amyloid plaques may be observed because arginine deiminase from Pseudomonas aeruginosa and peptidylarginine deiminase from bovine brain are inhibited by amyloid peptides that contain arginine (amyloid 1-42), and enhanced peptidylarginine deiminase activity is noted with free L-arginine [4,10,28]. Most bacteria in AD patients were not found in SPF model mice, strongly suggesting that studies of gut microbiota in aseptic mice do not tell the full story. Investigations in wild gut microbiota mice and long-term clinical studies are urgently needed to determine if these bacterial alterations are a cause or effect of AD. A recently report ensured that the important role of microglia and NLRP3 in ammasome activation in the pathogenesis of tauopathies and support the amyloid-cascade hypothesis in Alzheimer's disease, and demonstrated that neuro brillary tangles develop downstream of amyloid-beta-induced microglial activation [4,10,28], and NLRP3 is considered as an intracellular sensor that senses multiple microbial antigens and endogenous danger signals [30,31]. In this study, serum levels of the cytokines MIP-1α was altered after cohousing (Fig. 5B).
There were also changes in brain levels of in ammatory factors including, TREM2, IL-12α, IL-23, NF-κB p65, NLRP3and TLP4, (Fig. 6 and Fig. 7). The AD biomarkers Amyloid-β-A4, Tau, p-Tau, and APOE were upregulated. So, we sure that the gut microbiota appear to play an indispensable role in modulating the gut-brain axis and could be an important pathogenic factor of AD, and target on the microbiome-gut-brain axis will be an effective action. However, it is not clear if this is a useful diagnostic biomarkers as the AD gut microbiota could be distinguished from the healthy group but not the disease control groups, it perhaps due to the small sample size.
Tau hyperphosphorylation is associated with abnormal Aβ aggregation in AD, there are speci c temporal patterns of phosphorylated Tau in different parts of the brain [4,10]. Tau hyperphosphorylation plays a vital role in regulating synaptic function and maintaining cytoskeletal integrity [4,10,28,30]. In this study, phosphorylation at sites Ser404 and Ser416 were upregulated after cohousing, indicating that the AD gut microbiota enhances Tau protein hyperphosphorylation at multiple sites. There are many factors regulating Tau phosphorylation level, including oxidative stress, in ammation, neurogenesis, immune response, dysbacteriosis, infection and autophagy dysfunction. Tau phosphorylation is also affected by oxidative stress, endoplasmic reticulum protein folding dysfunction, and protein clearance ability decreases mediated by the proteasome and autophagy [4,10,28,30]. Normally, this abundant soluble protein can promote microtubule assembly and stability in axons, but this balance is upset when the Tau protein is hyperphosphorylated due to infection, metabolic disease, or chronic in ammation. Tau will be hyperphosphorylated and/or depolymerized depending on the activities of GSK-3β and the cell cycle protein-dependent kinase p25 [4,10,28]. As in Fig. 8A, the expression of GSK-3β was upregulated by AD feces extracts, and the autophagy was dysfunction.
Autophagy is a critical cellular process of internal degradation and recycling harmful or damaged components [31]. Autophagy dysfunction and tissue in ammation make people more susceptible to diseases, especially to intestinal diseases [32]. As a conservative serine/threonine protein kinase, mTOR is the junction of upstream pathways to regulate the cell growth, proliferation, movement, survival and autophagy [33], mTORC1 promotes the cell growth and metabolism, and inhibits autophagy by binding ULK1 complex [33]; and previous study have demonstrated that the leucine can activate the mTOR to regulate the autophagy and mitochondrial function [23]. In this study, we found that the levels of leucine in feces were reduced, while norleucine levels were increased in serum, expressions of mTOR and TORC1 were up-regulated in cohousing mice (Fig. 7A), and autophagy dysfunction (Fig. 7A), which indicate that AD-cohousing disorder the leucine metabolism, activate the mTOR to unregulated the autophagy.
Dysfunction of autophagy aggravates the accumulation of pathological products and in ammation levels (Figs. 6-7). Moreover, systemic in ammatory reactions caused by compounds secreted by bacteria promote oxidative stress, neuroin ammation, autophagy and/or neurodegeneration as demonstrated in previous studies [28]. So, we conclude that harmful microbes and/or their metabolites from APP/PS1 mice implanted in the newborn mice, caused metabolic imbalance, activated chronic in ammatory responses and affected autophagy and tau protein hyperphosphorylation (cover). Therefore, targeting the gut microbiota could be an effective treatment to slow AD progression. The role of the complex gut microbiota in AD still requires further investigation both at the community and/or strain level.
Multivariable-adjusted analyses showed that sphingomyelins and ether-containing phosphatidylcholines were altered in preclinical biomarker-de ned AD stages, whereas acylcarnitines and several amines, including the branched-chain amino acid valine and α-aminoadipic acid, changed in symptomatic stages [4,10,28,30,34]. Decreased neuronal glucose metabolism that occurs in AD brain could play a central role in disease progression [4,10,28,30,34]. And more evidences showed that amino acid oxidation can temporarily compensate for the decreased glucose metabolism, but eventually altered amino acid and amino acid catabolite levels likely lead to toxicities contributing to AD progression. Because amino acids are involved in so many cellular metabolic and signaling pathways, the effects of altered amino acid metabolism in AD brain are far-reaching [4,10,28,30,34]. In this study, we found that amino acid metabolism and lipid metabolism were imbalanced (Fig. 9, table S3), and we demonstrated that the leucine metabolism imbalance induced the autophagy dysfunction, which aggravates the accumulation of pathological products and in ammation levels, these also indicated that regulate the metabolism balance targeting on the gut-brain axis is important for AD prevention and treatment.
Dietary, microbial, and in ammatory factors modulate the gut-brain axis and in uence physiological processes ranging from metabolism to cognition [35]. Nutrients affect gut microbiota composition and the formation and aggregation of cerebral Aβ [35]. In a transgenic AD mouse model, AD pathology shifted gut microbiota composition toward an in ammation-related bacterial pro le, suggesting that these changes could contribute to disease progression and severity [7]. The gut microbiota has been shown to mediate the anti-epileptic effect of a ketogenic diet [7,35]. Metabolites of dietary tryptophan produced by micro ora control microglial activation, affect TGFα and vascular endothelial growth factor-β production, regulate transcription in astrocytes, and modulate CNS in ammation via the aryl hydrocarbon receptor, with implications for anxiety and depression [36]. The diet also impacts AD progression; germ-free mice demonstrated de cits in nonspatial and working memory tasks, as well as reduced hippocampal expression of brain-derived neurotrophic factor [37]. Uridine-and docosahexaenoic acid-containing diets could prevent rotenone-induced motor and gastrointestinal abnormalities associated with the pathogenesis of PD [37]. Altered gut microbiota composition has been associated with the onset of AD, which is characterized by the cerebral accumulation of amyloid-β brils [37]. It is therefore possible that modulation of the gut microbiome by speci c nutritional intervention may prove to be an effective strategy to prevent or reduce the risk of neurodegenerative disorders, such as PD and AD.

Conclusions
The gut microbiota structure of AD patients are different from control groups, which suggest that there is a certain correlation between gut microbiota and AD. Special pathogenic bacteria were found in AD patient samples, which might induce and/or accelerate disease processes. Gut microbiota from APP/SP1 mice changed structure of gut microbiota of newborn mice. The effect of changes in gut microbiota structure in cohousing mice, which may cause the leucine metabolism disorder and induces the mTOR mediates autophagic dysfunction, then aggravates in ammation levels and Tau protein phosphorylation.
Subsidiary experiments in BV2 cells shows the same trends. These ndings indicate that the metabolites of gut microbiota may induce and/or accelerate autophagy, in ammation, and neurodegeneration and participates in the process of AD by accelerating Tau protein phosphorylation.

Study Limitations
There are several strengths of this study, but there are also limitations. For example, we couldn't gure out which bacterial specie is associated with AD, or maybe many microorganisms with the whole action. And we have not perform a rescue test with single bacteria species, so we could not con rm direct contact between p-Tau and single bacteria, and we could not determine the speci c signaling pathway. Fecal transplants of targeted strains in germ-free mice with multi-omics studies will clarify interactions in the microbiome-gut-brain axis, and experiments in genetically altered mice could provide further evidence supporting our hypothesis. And long-term clinical monitoring are needed to screen dietary and nutritional interventions for AD based on gut microbiota.

Declarations
Ethical Approval and Consent to participate The animal protocols used in this work were approved by the Institutional Animal Care and Use committee of the Center of Laboratory Animals of the Guangdong Institute of Microbiology.

Consent for publication
Not applicable.

Availability of data and materials
Data collection and sharing for this project was funded by the National Natural Science Foundation of China.

Competing interests
The authors declare that they have no competing interests.

Funding
The present work was supported by the nancial support from the National Natural Science Foundation of China (81701086) Authors' contributions Chen DL and Wang Jian designed the study, carried out the computational analyses and wrote the manuscript; Guo YR, Qi LK and Tang XC collected animal physiological data and fecal samples and  Three-phase diagram of gut microbiota structure (at phylum level) in AD patients. Feces was collected from 48 AD patients, 28 disease controls, and 26 healthy controls. The 16S rRNA analysis showed signi cant differences for the abundance-based covered estimator (ACE) and Chao 1 indexes from the healthy groups (p < 0.05). (* denote p < 0.05, ** denote p < 0.01, *** denote p < 0.001) Figure 2 The gut microbiota composition of APP/SP1 mice. (A) The heatmap of gut microbiota at the genus level and Statistical chart of some bacterial of genus level. The gut microbiota composition of APP/PS1 mice using 16S rRNA sequencing, and the results revealed that the gut microbiota composition of APP/PS1 mice (n = 8) were very different from the normal C57 mice (n = 11). (* denote p < 0.05 , ** denote p < 0.01, *** denote p < 0.001) Figure 3 Colonization of gut microbiota analysis after AD-cohousing. (A) Alpha diversity analysis results; (B) Changes in the gut microbiota at the genus level and Statistical chart of some bacterial of genus level. After 2 weeks of pregnancy, the dams were placed with two 8-month-old APP/PS1. The mice were cohoused until 14 days after birth, and at rst and second month time points, the gut microbiota composition of intestinal contents were analyzed using 16S rRNA sequencing. S1N denote cohousing mice at rst month time point. S2N denote cohousing mice at second month time point. AD-cohousing denote cohousing mice at tenth month time point.
Page 28/39 Figure 4 LTP in hippocampal slices after cohousing. Ten months later, subsets of mice were randomly selected to perform standard eld potential recordings. (A)Repetitive stimulation (0.33 Hz) of Schaffer collaterals evoked fEPSPs in the hippocampal CA1 region. (B)The input-output relationship curves (C) Linear slopes revealed that compared to control mice (NC), the hippocampal CA1 neurons of AD-cohousing mice showed decreased excitability (p < 0.05, unpaired t test). (D) Synaptic short-term plasticity was measured using a PPF protocol in hippocampal slices. Statistical analyses from all slices demonstrated a signi cant inhibition of the ratio of P2/P1 at all tested P1 and P2 intervals in AD-cohousing mice compared to control mice of NC. (E)The plot recording time to normalized fEPSP slopes (baseline as 1) from pooled data showed impaired LTP induction (after theta-burst stimulation 0-10 min) and maintenance (after theta-burst stimulation 50-60 min) in all treated groups. (p < 0.05, unpaired t test) (G) The mean LTP induction results after theta-burst stimulation 0-10 min. (p < 0.05, unpaired t test). (F) The mean LTP induction results after theta-burst stimulation 50-60 min. (p < 0.05, unpaired t test). Data are presented as the means ± SD of more than 6 independent experiments. *p <0.05 and **p <0.01 vs. the model group by one-way ANOVA, followed by the Holm-Sidak test. (NC denote the control group, NS denote the AD-cohousing group) Figure 6