Meng-Fei Sun, Ying-Li Zhu, Zhi-Lan Zhou, Xue-Bing Jia, Yi-DaXu, Qin Yang, Chun Cui, Yan-Qin Shen

Abstract
Parkinson’s disease(PD)patients display alterations in gut microbiota composition. However, mechanism between gut microbial dysbiosis and pathogenesis of PD remains unexplored, and no recognized therapies are available to halt or slow progression of PD. Here we identified that gut microbiota from PD mice induced motor impairment and striatal neurotransmitter decrease on normal mice. Sequencing of 16S rRNA revealed that phylum Firmicutes and order Clostridiales decreased, while Thermal Cyclers phylum Proteobacteria, order Turicibacterales and Enterobacteriales increased in fecal samples of PD mice, along with increased fecal short-chain fatty acids (SCFAs). Remarkably, fecal microbiota transplantation (FMT) reduced gut microbial dysbiosis, decreased fecal SCFAs, alleviated physical impairment, and increased striatal DA and 5-HT content of PD mice. Further, FMT reduced the activation of microglia and astrocytes in the substantia nigra, and reduced expression of TLR4/TNF-α signaling pathway components in gut and brain. Our study demonstrates that gut microbial dysbiosis is involved in PD pathogenesis, and FMT can protect PD mice by suppressing neuroinflammation and reducing TLR4/TNF-α signaling. Keywords: Parkinson’s disease, microbial dysbiosis, gut-brain axis, short-chain fatty acids, glia, neuroinflammation, neurotransmitter, TLR4/TNF-α signaling

1.Introduction
Parkinson’s disease (PD) is the second most common neurodegenerative disorder clinically characterized by motor and non-motor symptoms. The classical motor symptoms include bradykinesia, resting tremor, rigidity and late postural instability, due to progressive death of dopaminergic neurons in the substantia nigra pars compacta (SNpc) (Jankovic, 2008). However,there is also a wide spectrum of non-motor manifestations involving gastrointestinal(GI) dysfunction,such as dysphagia,constipation and weight loss, which occur frequently in PD and may precede motor symptoms (Goldstein et al., 2011; Kim and Sung, 2015). The emerging role of gut microbiota in GI pathogenesis and bidirectional communication between the gut and brain in PD has received much attention. Growing evidence shows that alterations in gut microbiota composition and microbial metabolites may be involved in the pathogenesis and clinical phenotype of PD (Scheperjans et al., 2015a; Unger et 4al., 2016). For example, PD patients showing a significant reduction of Prevotellaceae in their feces, as compared to normal individuals. Moreover, a direct correlation 49 between the abundance of Enterobacteriaceae and severity of postural instability and gait difficulty has been found in PD patients (Scheperjans et al., 2015a). Another study also demonstrated that the colonic bacterial phylum Bacteroidetes and the bacterial family Prevotellaceae were reduced while Enterobacteriaceae were more abundant in PD patients (Keshavarzian et al., 2015).

Interestingly, gut microbiota from PD mice overexpressing “-synuclein (“-Syn) is required for motor deficits, 55 microglia activation and “-Syn pathology, and germ-free mice receiving fecal microbiota transplantation (FMT)from PD patients display increased motor dysfunction(Sampson et al., 2016). However, the evidence for gut microbial dysbiosis in animals is lacking, and how gut microbial dysbiosis contributes to PD pathogenesis remains unknown. Similar to PD patients (Petrov et al.,2017; Scheperjans et al., 2015b), we found that PD mice also display gut microbial dysbiosis, based on 16S rRNA sequencing. Remarkably, fecal samples from PD mice exhibit manifestations of gut microbial dysbiosis specifically with decreases in the phylum Firmicutes and order Clostridiales, and increases in the phylum Proteobacteria, order Turicibacterales and Enterobacteriales.00

To clarify the role of gut microbial dysbiosis in PD pathogenesis, we used a microbiota-targeted technique, named FMT, whereby infusion of feces (the entire gut microbiota) was delivered into the GI tract with gut microbial dysbiosis to re-establish a normal gut microbial community (Khoruts and Weingarden, 2014). Based on gut microbial dysbiosis, FMT is now applied for the treatment of autism, multiple sclerosis and other CNS diseases, possibly to animals were maintained (5 108 mice/cage, 12 h light/dark cycle) under pathogen-free conditions in temperature (24 ± hen PBS containing 20% sterile 118 glycerol (200 μl per mouse) by gavage for 7 days; and the MPTP+FMT group (n=15 119 mice), which was treated with MPTP by i.p injection for 5 days, then followed by FMT treatment (fecal microbiota from normal control mice) for 7 days. In addition, to test the hypothesis that gut microbial dysbiosis is necessary to promote PD pathology, we designed the “NS+PD-FMT” group (n=10 mice) which was treated with sterile normal saline (NS) for 5 days by i.p injection, then followed by PD-FMT treatment (fecal microbiota from PD mice) for 7 days.

Besides, two groups were designed to assess the potential effects of solvent and FMT, which named NS+PBS group (n=10 mice) and NS+FMT group (n=10 mice). The NS+PBS group mice got NS by i.p for 5 days then PBS containing 20% sterile glycerol by gavage for 7 days, and the NS+FMT group mice got NS by i.p for 5 days then FMT from normal mice for 7 days. MPTP treatment: C57BL/6 mice received i.p. injections of MPTP (30 mg/kg) in a volume of 10 mL/kg of body weight once daily for 5 days. FMT treatment: Fresh fecal pellets were collected from PD mice or healthy control mice, then diluted immediately with sterile PBS (1 fecal pellet/ml). Briefly, the stool was steeped in sterile PBS for about 15 min, shaken and then centrifuged at 1000 rpm, 4 °C for 5 min. The suspension was centrifuged at 8000 rpm, 4 °C for 5 min to get total bacteria, then filtered twice in PBS. The final bacterial suspension was mixed with an equal volume of 40% sterile glycerol to a final concentration of 20%, then stored at −80 ºC until transplantation (Hamilton et al., 2012). For each mouse, 200 μl of bacterial suspension (108 CFU/ml) (Schuijt et al., 2016) was transplanted to each of the recipient mice by gavage each day for consecutive 7 days. The dilution coating method was used for determination of microbial concentrations from mice by calculating colonies on solid medium under anaerobic incubation as previously described (Ubeda et al., 2013). On the 4th day of PBS or FMT treatment, the mice began behavioral training once per day for 3 days. The behavioral tests were performed on the first day after the last PBS or FMT treatment. Protocols of behavioral tests have been described in previous studies (Cao et al., 2017; Guo et al., 2015; Kuribara et al., 1977) .

Pole descent test: a 0.5 m long pole, 1 cm in diameter and wrapped with non-adhesive gauze to facilitate gripping, and with a spherical protuberance, 2 cm in diameter on top, was placed in the home cage. Mice were placed head-down on the top of the pole and their descent back into the home cage was timed. Timing began when the experimenter released the animal and ended when one hind-limb reached the home cage base. The test was performed three times for each animal, beginning at the day after the last treatment, and the average time was taken. Traction test: the forepaws of the mice were placed on a horizontal rope (diameter 5 mm) and observed for 10 seconds, while the hind limb placements were scored from 1 to 4, with the lowest score indicating the most severe deficit. Animals were assigned a score of 4 for gripping the rope with both hind paws, 3 for gripping the rope with one hind paw, 2 for gripping the rope with both front paws and 1 for gripping the rope with one front paw. The test was performed three times for each animal, beginning at the day after the last treatment, and the average score was taken. For collecting feces, normal control mice, MPTP-treated mice and FMT treated mice were placed individually in empty autoclaved cages and allowed to defecate freely in the morning after the day of last treatment. Once feces were formed of each mouse, they were collected immediately in individual sterile EP tubes on ice and then stored at −80 °C until next usage.

For obtaining fresh striatum and colon tissue, mice were deeply anesthetized by isoflurane and then received a transcardiac perfusion of ice-cold sterilized saline (20 ml per mouse). The tissues were quickly dissected and immediately stored at −80 °C. For immunofluorescence (IF), mice were transcardially perfused with PBS followed by 4% paraformaldehyde sequencing data was filtered according to the phred scores and the reads were trimmed if the average phred score in the window (5 bp in size, 1 bp step length) was less than 20. The processed reads that were shorter than 150 bp were removed. After trimming, the reads were assembled by Flash software (Magoc and Salzberg, 2011), and the reads that could not be assembled were discarded. Then sequences that contained ambiguous bases, more than 1 mismatch in their 5’ primers, or had a homopolymer over 8 or sequence length shorter than 200 bp were removed by Qiime (Caporaso et al., 2010). Only those high quality sequences without chimeras were used for further analysis. Uclust (Edgar, 2010) in Qiime was used to cluster the high quality sequences with a 97% similarity to obtain operational taxonomic units (OTUs). Then representative sequences of OTUs were annotated by comparing to the Greegenes database (13.5). Weighted principal coordinate analysis (PCoA) was performed using UniFrac analysis (Lozupone and Knight, 2005) to represent the similarity among the microbial communities. Heatmap and PCoA were performed on R (version 3.0.3), while Venn diagrams were implemented by Venn Diagram software. Fresh fecal pellets were homogenized in sterile deionized water (1 mg feces/10 μl ddH2O; 10%, w/v), then centrifuged at 13000 rpm, 4 °C for 5 min, to pellet the particulate matter.

The supernatants were collected and filtered through a 0.22 μm filter, and 1 μl of sample was injected into the column. Quantitation of fecal SCFAs (acetic acid, propionic acid, butyric acid and n-valeric acid) was performed on a GCMS-QP2010 Ultra (Shimadzu Co., Tokyo, Shimadzu) fitted with an Rtx-WAX column (30 m × 0.25 mm × 0.25 μm, Bellefonte, PA, USA). The column was first kept at 100 °C, and after 1 min, the column temperature escalated to 140 °C at a rate of 7.5 °C/min, and then to 230 °C at 60 °C/min. Helium was used as the carrier gas at a flow rate of 0.89 ml/min and the split ratio was 10:1. Temperatures of the injector, interface and ion source were 240, 250 and 220 °C, respectively. The mass spectrometer was operated in selected ion monitor scan mode, and the selected ions for acetic acid, propionic acid, butyric acid and n-valeric acid were 43, 74, 60 and 60 m/z. The external standard method was used to obtain the contents of target acids in samples. SCFAs mixtures consisting of the four acids were prepared at a series of concentrations of 5, 10, 20, 40, 60, 80 and 100 μg/ml to construct the calibration curves. The calibration curve samples were treated in a similar way as the study samples. chromatography (HPLC) Striatal DA, 5-HT and their metabolites, including dihydroxy-phenylacetic acid (DOPAC), homovanillic acid (HVA) and 5-hydroxyindoleacetic acid (5-HIAA) were quantified by HPLC with a fluorescence detector (Waters 2475), following previous methods with some modifications (Ahmed et al., 2014). The separation system (Waters 2695, U.S.A.) was used with an Atlantis T3 column (150 mm*4.6 mm, 5 μm, Waters).

The mobile phases were water, acetonitrile and 0.01 M phosphate buffer (adjusted topH 4 with phosphoric acid), using gradient elution. Briefly, the freshly isolated striatum was homogenized in 0.1 M perchloric acid (1mg:10μl) and the homogenate was centrifuged at 13,000 rpm, 4 °C for 10 min. Then, the supernatants were collected and filtered through a 0.22 μm filter and 25 μl of sample was injected into the column. In parallel, DA, 5-HT and their metabolites (DOPAC, HVA, 5-HIAA), and a hydrochloride standard (Sigma-Aldrich) solution were prepared freshly by diluting the stock solution in the mobile phase. The linearity range of DA was determined by serial concentrations of standard solution before detection. Total protein was extracted by homogenizing 20 mg tissue in RIPA buffer (Beyotime, China) with 1% phenylmethanesulfonyl fluoride (PMSF) (Solarbio, China). After centrifugation of the homogenate at 13,000 rpm, 4 °C for 5 min, protein concentration in the supernatant was determined with a BCA kit (Soliba, China). The supernatant was denatured by boiling for 5 min in SDS sample buffer. 40 μg of total protein were separated by 6%– 15% SDS-PAGE, blotted onto PVDF membranes, and then probed with the following antibodies: mouse anti-tyrosine hydroxylase (1:1000, MAB318, Millipore), rabbit anti-TLR4 (1:1000, 19811- 1-AP, Proteintech), rabbit anti-NF-κB (1:1000, #8242, Cell Signaling Technology), rabbit anti-TBK1 (1:2000, ab40676, Abcam) and mouse anti-GAPDH (1:1000, AF0006, Beyotime, China). Goat anti-rabbit IgG (1:1000, A0208, Beyotime, China) and goat anti-mouse IgG (1:1000, A0216, Beyotime, China) conjugated to horseradish peroxidase were used as secondary antibodies.

Protein bands were visualized by incubation with BeyoECL Plus (P0018, Beyotime, China) for 1 min and imaged by a Gel Image System (Tanon, 5200, China). Densitometry was performed by using Image J software. TNF-α concentration of striatum and colon was detected using a commercial ELISA kit (Boster, EK0527, China). The limit of detection is 15.6 pg/ml~1000 pg/ml. All experimental procedures were performed according to the manufacturer’s instructions. TNF-α concentration was expressed as pg/ml protein. Briefly, each mouse brain was cut into coronal slices and slices containing the SNpc were collected. Brain sections were immersed in 0.01 M sodium citrate buffer (pH 6.0) for antigen retrieval and washed twice in PBS. Then, brain sections were 281 incubated in PBS containing 0.2% v/v Triton X- 100, 0.02% w/v sodium azide and 5% v/v goat serum for 1 hour at 37 °C. Primary antibodies: mouse anti-tyrosine hydroxylase (1:1000, MAB318, Millipore), rabbit anti-GFAP (1:2000; Z033429, Dako, Denmark) and anti-Iba- 1 (1:1000, 019- 19741, Wako, Japan) were applied overnight at 4 °C. For detection of the primary antibodies, appropriate secondary antibodies, coupled to FITC-conjugated goat anti-mouse IgG (1:1000, A0568, Beyotime, China) or CY3-conjugated goat anti-rabbit IgG (1:1000, A0516, Beyotime, China) were used. Samples were covered with mounting medium (P0126, Beyotime, China) and examined with an epifluorescence microscope (Nikon eclipse 80i, Nikon, Japan). For each animal, brain slices containing major part of SN from bregma -2.92 mm to -3.52 mm, 6 representative sections (space 100 μm between) were chose to be doubly stained with tyrosine hydroxylase (TH) and Iba- 1 or GFAP respectively. The cells positive for TH, Iba- 1 and GFAP were counted in the right SN and the positive cells evaluation was used by the Image J analyzer(NIH,U.S.A.).The above-mentioned 6 representative sections were quantified for each animal, and each group contains 6 animals. These data are present as mean number of cells per section per group. SPSS 22.0 software was used for data analysis. Statistical analysis was conducted by one-way ANOVA with an LSD post hoc assay. Data are presented as mean ± s.e.m (standard error of the mean), and P<0.05 was set as the threshold for significance (*P<0.05, **P<0.01, ***P<0.001). All data were displayed using Graph Pad Prism version 5.01 (Graph Pad, Inc, La Jolla, U.S.A.).

Results
dysfunction commonly occurs in PD.To assess the potential neuroprotective effects of FMT on motor function in MPTP-induced PD mice, we subjected mice to a pole test to measure the total descent time for evaluation of bradykinesia, and a traction test to measure the traction score for evaluation of muscle strength and equilibrium. MPTP+PBS mice exhibited motor dysfunction, including longer pole descent time (p<0.001 vs. normal controls) and lower scores in the traction test (p<0.001 vs. normal controls) compared with normal control mice. However, FMT-recipient mice displayed significantly improved performance in pole descent test (p<0.001 vs. the MPTP+PBS group) and traction test (p<0.001 vs. the MPTP+PBS) compared with PBS treated mice (Fig.1A and 1B). These results suggest that FMT improves the motor functions in PD model mice. sess the solvent effect and potential effects of gut microbiota from normal mice and PD mice on motor function in normal mice, pole descent test and traction test were also tested in NS+PBS group, NS+FMT group and NS+PD-FMT mice, MPTP+PBS group was set as PD model control. Total descent time was significantly longer in MPTP+PBS mice and NS+PD-FMT mice compared with NS+PBS mice and NS+FMT mice (Fig. S1A), respectively. Consistently, significant decrease in traction test score was observed in MPTP+PBS mice and NS+PD-FMT mice compared with NS+PBS mice and NS+FMT mice (Fig. S1B). There were no differences between NS+PBS group and NS+FMT group including pole descent test and traction test. These behavioral test results suggest that solvent or FMT has no effect on motor function of normal mice.

FMT reduces motor dysfunction of PD mice while gut microbiota from PD mice promotes motor dysfunction of normal mice. (A) Pole descent test: time to descend the pole, as a measure of bradykinesia. Total descent time was significantly longer in MPTP+PBS mice compared with normal control mice, and shorter in MPTP+FMT mice compared with MPTP+PBS mice. (B) Traction test: traction reflex score for evaluation of muscle strength and equilibrium. The traction reflex score was significantly lower in and MPTP+PBS mice compared with normal control mice, and higher in MPTP+FMT mice compared with MPTP+PBS mice. Statistical comparison by one-way ANOVA with post hoc comparisons of LSD; Data represent the means ± s.e.m; ***P<0.001. n=12 mice per group. To evaluate potential neuroprotective effects of FMT on brain function, the striatal neurotransmitters DA and 5-HT, and their metabolite (DOPAC, HVA and 5-HIAA) concentrations were measured by fluorescence detection following HPLC. As expected, striatal DA levels were dramatically decreased by 75.0% in MPTP+PBS mice compared with normal control mice (p<0.001 vs. normal controls), but increased <0.05 vs. MPTP+PBS group) (Fig. 2A). Similarly, DOPAC and HVA were also decreased by 17.5% and 45% in MPTP+PBS mice compared with control mice (p<0.05 and p<0.001 vs. normal controls, respectively). However, in MPTP+FMT mice, DOPAC and HVA were increased by 27.6% and 66.7% compared with the MPTP+PBS mice (p<0.01 and p<0.001 vs. MPTP+PBS group, respectively) (Fig. 2B and 2C).

Similar to DA, 5-HT and its’ metabolite 5-HIAA were also decreased by 50.0% and 44.8% in MPTP+PBS mice compared with normal control mice (p<0.001 and p<0.05 vs. normal controls, respectively), and a significant increase by 66.7% and 125% in MPTP+FMT mice compared with MPTP+PBS mice (p<0.01 and p<0.001 vs. MPTP+PBS group, respectively) (Fig. 2D and 2E). These results suggest that FMT participates in the metabolism of brain neurotransmitters by suppressing the MPTP-induced decline of striatal DA, 5-HT and their metabolites. To assess the solvent effect and potential effects of gut microbiota from normal mice and PD mice on striatal neurotransmitters of normal mice, DA and 5-HT, and their metabolite (DOPAC, HVA and 5-HIAA) concentrations were also measured in NS+PBS group, NS+FMT group and NS+PD-FMT group, MPTP+PBS group was set as PD model control. Decreased DA levels (Fig. S2A), DOPAC levels (Fig. S2B), HVA levels (Fig. S2C), 5-HT levels (Fig. S2D) and 5-HIAA levels (Fig. S2E) in MPTP+PBS and NS+PD-FMT group were found when compared with NS+PBS and NS+FMT groups respectively, and there were no differences between NS+PBS group and NS+FMT group. Clearly, solvent has no effect on striatal neurotransmitters of normal mice, and PD-FMT reduces neurotransmitters significantly in normal mice. FMT treatment of PD mice inhibits the MPTP-mediated reduction in brain neurotransmitters and TH. (A) DA content in striatum shows higher DA expression in the MPTP+FMT group compared with the MPTP+PBS group. (B) DOPAC content in striatum shows the MPTP-mediated decrease of DOPAC expression was reversed by FMT. (C) HVA content in striatum shows the MPTP-mediated decrease of HVA expression was inhibited by FMT.(D) 5-HT content in striatum shows the MPTP-mediated decrease of 5-HT expression was inhibited by FMT. (E) 5-HIAA content in striatum shows the MPTP-mediated decrease of 5-HIAA expression was inhibited by FMT. DA, 5-HT and their metabolites were analyzed by HPLC. DA, 5-HT and their metabolites were analyzed by HPLC.

Statistical comparison by one-way ANOVA with post hoc comparisons of LSD; Data represent the means ± s.e.m; *P<0.05, **P<0.01, ***P<0.001. n=8 mice per group. To examine the effect of FMT on survival of dopaminergic neurons in the SN, and to verify whether TH expression corresponded with DA levels, we characterized TH expression by western blot analysis and IF staining in the SN. IF staining in the SN revealed a significant loss of TH+ dopaminergic neurons by 31.6% in MPTP+PBS mice compared to normal control mice (p<0.001 vs. normal controls) that was partially rescued 32.6% by FMT (p<0.01 vs. MPTP+PBS group) (Fig. 3A and 3B). Western blot analysis of striatal tissue showed that TH expression in MPTP+PBS mice was lower by 68.6% compared to normal control mice (p<0.01 vs. normal controls), and FMT inhibited the reduction by 82.8% of TH expression in PD mice (p<0.05 vs. the MPTP+PBS group) (Fig. 3C and 3D). These results demonstrate that the neuroprotective effects ofFMT involve suppression of dopaminergic neuron loss in SN and decreased striatal TH expression in MPTP-induced PD mice. Fig. 3. FMT treatment of PD mice rescued the MPTP-mediated reduction in brain dopaminergic neurons and TH. (A) Representative IF staining for TH (dopaminergic neuron marker) in the SN. Scale bar is 50 μm. (B) Quantitative analysis of the number of TH-positive cells in the SN. (C) Representative western blot of striatal TH expression. (D) The intensity of bands was quantified with Image J software and quantitative data for TH following normalization to GAPDH). Statistical comparison by one-way ANOVA with post hoc comparisons of LSD. Data represent the means ± s.e.m; *P<0.05, **P<0.01, ***P<0.001. n=6 mice per group. To test our hypothesis that gut microbial dysbiosis also exists in PD mice (MPTP+PBS group) and FMT can improve the gut microbial dysbiosis in PD mice by affecting the microbiota composition, we performed relative analysis based on alpha and beta diversity measures, which provided a holistic view of gut microbiota and focus on abundance, diversity and distribution of gut microbiota.

The alpha diversity indices of Chao1, representative of microbial diversity, tended to be higher in normal control (C) mice and MPTP+FMT (F) mice, compared to MPTP+PBS (M) mice (Fig. 4A), but there were no significant differences between the three groups, suggesting the number of samples need to be expanded. Beta diversity was further evaluated with UniFrac-based PCoA. Strikingly, using analysis of similarities (ANOSIM) of normal control, MPTP+PBS and MPTP+FMT groups, samples were clustered by subject (p<0.01), indicating differences in microbial composition of PD mice, control mice and FMT-recipient mice (Fig. 4B). Also, other measures of alpha diversity indices including Shannon diversity and PD Whole tree was similar with Chao- 1 (Fig. S3A and S3B). The differences in microbial communities between PD mice and normal mice suggested the presence of gut microbial dysbiosis in PD mice. To explore the abundance and distribution in gut microbiota in the three groups, and investigate the potential bacterial groups responsible for microbial dysbiosis, we identified a number of altered orders and phyla. Only those bacterial phyla and orders that averaged more than 1% of the relative abundance across all samples are displayed. The heatmap revealed a significant difference in relative abundance across the groups at the phylum level (Fig. 4C). A significant decrease was observed in the abundance of the phylum Firmicutes principally at the phylum level in MPTP+PBS mice compared to control mice. Interestingly, there were obvious increases in Firmicutes in FMT-recipient mice (Table 1). The abundance in phylum Proteobacteria was strikingly increased in MPTP+PBS mice compared to normal control mice, but comparatively lower in MPTP+FMT mice (Table 1).

At the order level, the overall composition of gut microbiota was dominated by the order Clostridiales (Fig. 4D). Decreased Clostridiales and increased Turicibacterales and Enterobacteriales were observed in PD mice compared to normal control mice, and FMT altered near the level of normal control (Table 2). Not only at the phylum and order level, significant difference of gut microbial was also obviously showed at genus level among three groups (Fig. S3C). Together, these data indicate that PD mice have gut microbial dysbiosis, and FMT can modulate the microbiota compositions to improve gut microbial dysbiosis in PD mice. Analysis of bacterial 16S rRNA from gut microbiota of PD mice shows FMT reduces PD microbial dysbiosis. (A) Analysis of alpha diversity-predicted diversity of gut microbiota by Chao- 1 estimator analysis. (B) PCoA plots based on weighted UniFrac metrics of gut microbiota where samples of mice from different groups are highlighted with different colors. Principal components (PCs) 1 and 2 explain 36.38% and 14.41% of the variance, respectively. The position and distance of data points indicates the degree of similarity in terms of both the presence and relative abundance of bacterial taxonomies. (C) Heatmap based on the Bray-Curtis distance analysis about relative abundance of gut microbiota at the phylum level in three groups. (D) Distribution of relative abundance of gut microbiota at the order level observed in the three groups. Each color represents an order, and the height of the column represents the abundance of reads. Only those bacterial orders and phyla that comprised on average more than 1% of the relative abundance across all samples are displayed. “C” represents the normal control group, “M” represents the MPTP+PBS group, “F” represents the “MPTP+FMT” group.

All other unassigned and classified OTUs belonging to phyla and orders comprising less than 1% of the total abundance are represented as Others/Unassigned. n=7 mice in normal control and MPTP+PBS groups and n=8 mice in MPTP+FMT group. ficant changes of orders >1% present in different groups. “C” represents the normal control group (n=7 mice), “M” represents the MPTP+PBS group (n=7 mice), and “F” represents the “MPTP+FMT” group (n=8 mice). Statistical comparison by one-way ANOVA with post hoc comparisons of LSD. Data represent the ± SEM; *P<0.05, **P<0.01, ***P<0.001. Among the potential factors regulating the microbiota-gut-brain axis, microbial metabolites SCFAs may be the major mediators. We detected the fecal SCFAs concentrations in different groups (Table 3). Among fecal SCFAs (acetic acid, propionic acid, butyric acid and n-valeric acid) detected by GC-MS, acetic acid was changed most, which increased by 260.6% in MPTP+PBS mice compared with normal control mice. Interestingly, FMT decreased fecal acetic acid by 82.7% concentrations of PD mice (Fig. 5A). Propionic acid was also increased but changed least among four acids, which increased by 70.0% in MPTP+PBS mice compared with normal control mice.

Instead, FMT decreased fecal propionic acid by 59.1% concentrations of PD mice (Fig. 5B). Similar to acetic acid, butyric acid and n-valeric acid were also increased in MPTP+PBS mice by 118.3% and 90.3% compared with normal control mice, while FMT significantly decreased fecal butyric acid and n-valeric acid by 72.0% and 65.4% concentrations of PD mice (Fig. 5C-5D). Given the evidence that SCFAs are sufficient to promote microglia-mediated neuroinflammation, based on morphology indicative of increased microglia activation (Sampson et al., 2016), we sought to determine whether brain glia are activated when fecal SCFAs increase abnormally. Thus, we measured the interactional activities between dopaminergic neurons and brain glia, including microglia and astrocytes, in the SN by IF staining. Double IF staining for TH (dopaminergic neuron marker) and GFAP (astrocyte marker) revealed the presence of a higher number of activated astrocytes by 91.4% in MPTP+PBS mice compared with normal control mice (P<0.001 vs. normal controls), and FMT significantly decreased the number of activated astrocytes around dopaminergic neurons by 37.3% on MPTP+PBS mice (p<0.001 vs. the MPTP+PBS group) (Fig. 6A and 6B). Similarly, co-expression of TH with Iba- 1 (marker for microglia) showed that, along with the loss of TH+ neurons, more microglia in the SN were activated by 125.8% in MPTP+PBS mice compared with control mice (p<0.001 vs. the normal controls), but in the MPTP+FMT group, activated microglia decreased by 45.3% compared with MPTP+PBS mice (p<0.001 vs. the MPTP+PBS group) (Fig. 6C and 6D). Fecal SCFAs alterations in mice. (A) Fecal acetic acid content. The MPTP-mediated increase of acetic acid expression was inhibited by FMT. (B) Fecal propionic acid content. The MPTP-mediated increase of propionic acid expression was inhibited by FMT. (C) Fecal butyric acid content.

The MPTP-mediated increase of butyric acid expression was inhibited by FMT. (D) Fecal n-valeric acid content. The MPTP-mediated increase of n-valeric acid expression was inhibited by FMT. SCFAs were analyzed by GC-MS. Statistical comparison by one-way ANOVA with post hoc comparisons of LSD. Data represent the means ± s.e.m; *P<0.05, **P<0.01, ***P<0.001. n=6 mice per group. SCFAs concentrations in different groups. “C” represents the normal control group, “M” represents the MPTP+PBS group, and “F” represents the “MPTP+FMT” group. Statistical comparison by one-way ANOVA with post hoc comparisons of LSD. Data represent the means ± SEM; *P<0.05, .01, ***P<0.001. n=6 mice per group. marker; red) in the SN. The white arrows indicated the GFAP-positive astrocytes in SN. Scale bar is 50 μm. (B) Quantitative analysis of the number of activated astrocytes in each group. Activation of astrocytes in the SN was lower in the MPTP+FMT group compared to the MPTP+PBS group. (C) Double IF staining of TH and Iba- 1 (microglia marker; red) in the SN. The white arrows indicated the IBA- 1-positive microglia in SN. Scale bar is 50 μm. (D) Quantitative analysis of the number of activated microglia in each group, and activation of microglia in the SN was inhibited in the MPTP+FMT group compared to the MPTP+PBS group. Statistical comparison by one-way ANOVA with post hoc comparisons of LSD. Data represent the means ± s.e.m; ***P<0.001. n=6 mice per group. TLR4/TBK1/NF-κB/TNF-α signaling pathway To further explore the molecular interactions between gut microbial dysbiosis and neuroinflammation in PD, we characterized colonic and striatal expression of TLR4, TBK1 and NF-κB by western blot, and TNF-α by ELISA. First, we observed the expressions were higher by 112.1%, 98.1%, 82.6% and 38.2% of TLR4, TBK1 and NF-κB, as well as TNF-α in the colon of MPTP+PBS mice (p<0.01, p<0.01, p<0.05 and p<0.05 vs. normal controls, respectively), whereas MPTP+FMT mice displayed decreased expressions of TLR4, TBK1 and NF-κB, as well as TNF-α by 40.0%,31.6%, 45.5% and 37.5% in the colon compared with MPTP+PBS mice (p<0.05, p<0.05, p<0.05 and p<0.01 vs. MPTP+PBS group, respectively) (Fig. 7A-7E). To verify the bidirectional communication between gut and brain, we also showed that TLR4/TBK1/NF-κB pathway was involved in brain neuroinflammation.

Significantly, striatal TLR4, TBK1, NF-κB and TNF-α expressions were higher by 111.9%, 270.0%, 72.8% and 12.4% in MPTP+PBS mice (p<0.01, p<0.01, p<0.05 and p<0.05 vs. normal controls, respectively), while MPTP+FMT mice displayed decreased expressions of TLR4, TBK1, NF-κB and TNF-α by 52.2%, 64.7%, 38.6% and 16.9% in the striatum compared with MPTP+PBS mice (p<0.01, p<0.01, p<0.05 and p<0.01 vs. the MPTP+PBS group, respectively) (Fig. 7G-7J). In summary, these results suggest that the TLR4/TBK1/NF-κB/TNF- α signaling pathway is involved in gut inflammation and neuroinflammation, and further that FMT may improve gut microbial dysbiosis to alleviate gut inflammation and neuroinflammation by suppressing this signaling pathway. FMT alleviates gut inflammation and neuroinflammation by suppressing the TLR4/TBK1/NF-κB/TNF-α signaling pathway. (A) Representative western blot of colonic TLR4, TBK1 and NF-κB expression. (B) MPTP-induced high expression of colonic TLR4 was inhibited by FMT. (C) MPTP-induced high expression of colonic TBK1 was inhibited by FMT. (D) MPTP-induced high expression of colonic NF-κB was inhibited by FMT. (E) MPTP-induced high expression of colonic TNF-α was inhibited by FMT. TNF-α expression was quantified by ELISA. (F) Representative western blot of striatal TLR4, TBK1 and NF-κB expression. (G) MPTP-induced high expression of striatal TLR4 was inhibited by FMT. (H) MPTP-induced high expression of striatal TBK1 was inhibited by FMT. (I) MPTP-induced high expression of striatal NF-κB was inhibited by FMT. (J) ELISA analysis of striatal TNF-α expression. MPTP-induced high expression of striatal TNF-α was inhibited by FMT. Statistical comparison by one-way ANOVA with post hoc comparisons of LSD. Data represent the means ± s.e.m; *P<0.05, **P<0.01. n=6~8 mice. For all western blot results, intensity of bands was quantified with Image J software, and band intensities were normalized to GAPDH.

Discussion
Given the recent evidence that gut microbial dysbiosis occurs in PD patients (Bedarf et al., 2017; Hill-Burns et al., 2017; Li et al., 2017), we also verified the composition differences of gut microbiota in a murine model of PD. Strikingly, our demonstration that gut microbial dysbiosis in PD mice involves decreases in phylum Firmicutes and order Clostridiales, and increases in phylum Proteobacteria, order Turicibacterales and Enterobacteriales increase, is consistent with observations in human subjects with PD, which have shown increased abundance of Proteobacteria (Keshavarzian et al., 2015; Unger et al., 2016) and Enterobacteriales (Unger et al., 2016). It is known that increases in proteobacteria can be a consequence of gut inflammation (Nagalingam et al., 2011). In particular, the abundance of Enterobacteriaceae has proven to be positively related to the severity of postural instability & gait difficulty in PD (Scheperjans et al., 2015a), a correlation also shown by our mouse model. These results suggest that gut microbial dysbiosis involving specific microbes maybe involved in PD pathogenesis and clinical manifestations. Recently, the critical role of gut microbiota and microbial metabolites in PD pathogenesis has received increasing attention (Houser and Tansey, 2017). We use PD model by systemic administration of the neurotoxin MPTP to explore the potential role of gut microbiota in PD. After crossing the blood brain barrier, MPTP is converted into its active metabolite MPP+ in glial cells by the enzyme monoamine oxidase B (MAO-B).

MPP+enters dopaminergic neurons through dopamine transporter (DAT) where it binds to complex I and impairs mitochondrial function. This stimulates the overproduction of free radicals which causes oxidative stress and ultimately leads to the activation of the cell death pathways (Dauer and Przedborski, 2003). Here, we show that gut microbiota from PD mice induces a hallmark of PD motor dysfunctions and neurotransmitter loss in healthy mice. This suggests that gut microbiota not only affects motor symptoms, but also brain function in PD. Consistent with a previous study showing that gut microbiota from PD patients promotes enhanced motor impairment in mice (Sampson et al., 2016), our results show that mice receiving gut microbiota from PD mice also display motor dysfunctions as well as decreases of striatal neurotransmitters. Levels of DA and the DA metabolite DOPAC as well as 5-HT and 5-HIAA are decreased in NS+PD-FMT mice, indicating that PD gut microbiota, but normal gut microbiota impacts brain neurotransmitters expressions. Conversely, PD mice, receiving fecal microbiota from normal mice, display significant recovery of DA, 5-HT and their metabolites in striatum, along with recovery of motor function. DA is an important neurotransmitter responsible for regulating balance and movement, and the perturbations of DA signaling are implicated in the pathogenesis or exploited in the treatment of PD (Tritsch and Sabatini, 2012). Our results indicate that gut microbiota can influence striatal DA concentration and FMT can suppress the loss of dopaminergic neurons in SN. 5-HT is a key neurotransmitter and signaling molecule in the regulation of mood at the level of the CNS and has been implicated in visceral hypersensitivity in the GI tract (Mayer, 2011).

It has been reported that gut microbiota constantly controls tryptophan metabolism of the host by kynurenine pathway, thereby simultaneously reducing the fraction available for serotonin synthesis and increasing the production of neuroactive metabolites (Heijtza et al., 2011; O’Mahony et al., 2015). Consistent with this, our data indicate that gut microbiota from PD promotes a decrease in striatal 5-HT concentration. The cross talk of serotonergic system between the gut and brain might explain these differences. Currently, there are no ideal therapies for PD. Based on the gut microbial dysbiosis in PD Biochemical alteration patients and mice, our data suggest that reducing gut microbial dysbiosis by FMT has a neuroprotective effect on PD. With a view to manipulate gut microbiota-host interactions, FMT is an exciting potential strategy not only for application in GI disease, such as Clostridium difficile infection (CDI) (Konturek et al., 2016) and inflammatory bowel disease (IBD) (Cammarota et al., 2016), but also for the treatment of CNS diseases, such as autism spectrum disorders (ASD) (Kang et al., 2017). Intriguingly, in our study, FMT not only alleviates motor impairment of PD mice, but also improves the dominant microbial population in PD mice, such as phylum Firmicutes and order Clostridiales. Moreover, we show FMT can rescue the decreases of striatal TH, DA, 5-HT and their metabolites, that occur in PD, while FMT has no side effect on behavioral functions and neurotransmitters on normal mice. These results suggest that FMT exerts a neuroprotective effect by improving gut microbial dysbiosis, resulting in reduced motor impairment and elevated of DA levels, and reduced loss of dopaminergic neurons, as well as restoration of 5-HT synthesis. In previous studies, after FMT, the microbes in recipients were able to be accurately identified for 3-6 months even up to 2 years (Kumar et al., 2017).

As a novel treatment, the safety and efficacy ofFMT has been proved in CDI patients and there were no other severe adverse events directly attributable to FMT (Hefazi et al., 2017). However, other potential effects of FMT on PD mice should be explored more. How gut microbial dysbiosis contributes to PD is not clear. Among the potential factors regulating the microbiota-gut-brain axis, microbial SCFAs may be one of major mediators. The study has revealed that differences in fecal SCFA ratios between PD patients and healthy controls, including elevated relative concentrations of butyrate consistent with the observed altered compositions of gut microbiota, implicate a potential role between SCFAs and gut microbiota in PD (Unger et al., 2016). In addition, SCFAs have been recognized as potential mediators of gut microbiota on intestinal immune function and regulation of inflammation (Vinolo et al., 2011). SCFAs are not only trafficked into the serum, but also are capable of crossing the blood brain barrier (BBB) via monocarboxylate transporters to impact the physiology of cells in the CNS such as maturation of microglia, for which mice deficient for the SCFA receptor FFAR2 mirrored microglia defects found under GF . While being fed a mixture of acetate, propionate and butyrate, the Zenidolol three most abundant SCFAs, PD mice display aggravated “-Syn-mediated neuroinflammation, including activation of microglia and motor symptoms (Sampson et al., 2016), suggesting a link between SCFAs and microglia-mediated neuroinflammation. Activation of microglia and astrocytes is responsible for progressive inflammation and neuronal cell death, giving rise to further inflammation and neurodegeneration in the SN-striatum system (Booth et al., 2017; Depboylu et al., 2012; Dutta et al., 2012). Our study also demonstrates that the higher fecal SCFAs concentrations, in PD mice, correlates with higher levels of activated striatal glia, including microglia and astrocytes.

Conversely, we observe a lower content of fecal SCFAs along with the lower activation of microglia and astrocytes in the SN of MPTP+FMT mice, compared with MPTP+PBS mice. The above-mentioned studies suggest that fecal SCFAs may contribute to the over activation of microglia and astrocytes in the SN, and FMT may exert a neuroprotective effect via suppression of glia cell activation and lowered expression of fecal SCFAs. Alterations in gut microbiota have been previously suggested, in Gulf War illness, to activate TLR4, in the small intestine and brain, to explain the intestinal inflammation and neuroinflammation observed (Alhasson et al., 2017). Furthermore, TLR4 deficiency was proved to attenuate up-regulation of AP- 1 in TH-positive neurons and astrocytes in the SN of MPTP-treated TLR4-KO mice, which suggested that TLR4 pathway may play an important role in PD pathogenesis (Zhao et al., 2016). Microbial dysbiosis in the gut may result in a leaky gut and release of microbial toxins, such as LPS, into the circulation, ultimately reaching the brain, to impact neurological function. Proteobacteria, shown here to be increased in PD mice, is a major gram-negative phylum in mice (Cheng et al., 2013), and is comprised of bacteria with LPS as the major cell wall component, which further stimulates the inflammatory response (Elinav et al., 2011).

Multiple mechanisms and pathways through which the brain and gut interact and influence host behavior involve the sympathetic and parasympathetic branches of autonomic nervous system, as well as the neuroendocrine and immune systems (Grenham et al., 2011). Based on the bidirectional communication between gut and brain, the release of pro-inflammatory factors, such as TNF-α through activation of TLR4/TBK1/NF-κB signaling pathway in the gut may induce gut immune cells activation which may release some inflammatory factors and then cross BBB to brain glia cells to cause neuroinflammation. Taken together, our findings establish that gut microbial dysbiosis is involved in the pathogenesis of PD,via the microbiota-gut-brain axis, whereby the gut microbiota in PD mice releases microbial components that impact gut inflammation and neuro inflammation. Further, we demonstrate that FMT can reverse gut microbial dysbiosis and result in neuroprotection, possibly by reducing activation of brain glia through lowering activity of the TLR4/TBK1/NF-κB/TNF-α signaling pathway in both gut and brain. Future studies may also reveal additional molecular markers that could be combined with gut microbial dysbiosis information to enhance the sensitivity and specificity of