Gut microbiota regulates autism-like behavior by mediating vitamin B6 metabolism in EphB6-deficient mice

Background Autism spectrum disorder (ASD) is a developmental disorder with limited effective pharmacological treatments for the core autistic symptoms so far. Increasing evidences, especially the clinical studies in ASD patients, suggest a functional link between gut microbiota and development of ASD. However, the mechanisms linking gut microbiota and brain dysfunctions (gut-brain axis) in ASD are still not well-established. With genetic mutations and down-regulated expression in patients with ASD, EPHB6 , which is also important in homeostasis of gut, has been generally considered to be a candidate gene for ASD. Nonetheless, the role and mechanism of EPHB6 involved in regulating gut microbiota and development of ASD have been unclear. Results Here, we found deletion of EphB6 induced autism-like behavior and disturbed gut microbiota in mice. More importantly, transplanting fecal microbiota from EphB6-deficient mice resulted in autism-like behavior in antibiotics-treated C57BL/6J mice. Meanwhile, transplanting fecal microbiota from wild-type mice ameliorated autism-like behavior in EphB6-deficient mice. At the metabolic levels, disturbed gut microbiota led to vitamin B6 and dopamine defects in EphB6-deficient mice. At the cellular levels, excitation/inhibition (E/I) imbalance in medial prefrontal cortex was induced by gut microbiota-mediated defects of vitamin B6 metabolism in EphB6-deficient mice. Conclusions Our study uncovers a key role for gut microbiota in regulation of autism-like social behavior by mediating vitamin B6 metabolism, dopamine synthesis and E/I balance in EphB6-deficient mice, suggesting new strategies for understanding and treatment of ASD.

mice clustered differently from 4-week-old WT mice (Additional file 1: Figure S2d), while gut microbiota of 3-week-old WT and KO mice clustered similarly (Additional file 1: Figure S2c). At the same time, 4-week-old KO mice, but not 3-week-old KO mice, showed increased self-grooming and decreased interest in social odor compared with even-aged WT mice (Additional file 1: Figure S2e-g).
These results further implied the possible relation between the abnormal behavior and gut microbial dysbiosis in mice with deletion of EphB6.
Transplantation of fecal microbiota from EphB6-deficient mice caused autism-like behavior in SPF C57BL/6J mice ASD is generally considered to be a neuro-developmental disorder, postnatal developmental disorder can also cause autism in patients [27], and postnatal mutation of Nrxn1 in neurons led to autism-like behavior in mice [28]. Also gut microbiota of ASD patients could induce autism-like behavior in mice [7]. So, to study the relation between gut microbial dysbiosis and autism-like behavior in mice with deletion of EphB6, we gavaged the fecal microbiota from 8-week-old male WT or KO mice to 3-weekold SPF male C57BL/6J mice for a week (Fig. 2a). Three weeks after the gavage of fecal microbiota, gut microbial composition in SPF C57BL/6J mice treated with fecal microbiota from WT and KO mice was different (Fig. 2b-d). More interestingly, C57BL/6J mice with the gastric perfusion of fecal microbiota from KO mice displayed increased self-grooming ( Fig. 2e) and decreased social behavior ( Fig. 2f-h) compared with control mice. While in open field test and elevated-plus-maze test, the two groups of mice behaved similarly (Additional file 1: Figure S2h-j). Furthermore, we gavaged orally the suspending solution of fecal microbiota from WT or KO mice to antibiotic-pretreated SPF male C57BL/6J mice. After pretreatment with antibiotics for 5 days, fecal microbiota of 8-week-old male WT or KO mice was gavaged orally to 3-week-old SPF male C57BL/6J mice for 5 days (Fig. 2j). About 2 weeks after fecal microbial colonization, similarly, we found gut microbiota of SPF C57BL/6J mice treated with fecal microbiota from KO mice clustered differently from control mice (Fig. 2k-m). Then we found C57BL/6J mice with gastric perfusion of fecal bacteria from KO mice showed increased selfgrooming ( Fig. 2n) and decreased social behavior (Fig. 2o-r). Also, the two groups of mice behaved similarly in open field test and elevated-plus-maze test (Additional file 1: Figure S2k-m). What's more, fecal microbiota from 4-week-old KO mice, but not 3-week-old, induced increased self-grooming and social deficits in 3-week-old SPF C57BL/6J mice compared with C57BL/6J mice gavaged with fecal microbiota from even-aged WT mice (Additional file 1: Figure S2n-s). Collectively, fecal microbiota from EphB6-deficient mice caused more self-grooming and impaired social behavior in C57BL/6J mice.
Then we wondered whether gut microbiota still played the role in autism-like behavior in adult mice.
To begin, we gavaged orally a mixture of antibiotics to 6-week-old male SPF C57BL/6J mice for a week. And we found antibiotic treatment disrupted the gut microbiota greatly and induced decreased self-grooming and social deficits in young adult C57BL/6J mice (Fig. 3a-i). These results indicated us that gut microbiota was related with autism-like behavior even in adult mice and different gut microbiota probably contributed to different behaviors, such as self-grooming and social behavior.
Then, we gavaged the fecal microbiota from 8-week-old male WT or KO mice directly to 6-week-old SPF male C57BL/6J mice for a week. And we found fecal microbiota from KO mice also induced disturbed gut microbiota, more self-grooming and social deficits in adult C57BL/6J mice ( Fig. 3j-r).
Basically, our results indicated the important role of gut microbiota in autism-like behavior, even in adult mice.

Transplantation of fecal microbiota from wild-type mice ameliorated autism-like behavior in adult EphB6-deficient mice
Until now, there has no studies focusing on the effectiveness of microbiota transplantation in adult ASD patients. Then we gavaged orally the fecal microbiota from 8-week-old male WT mice to 8-weekold KO mice for a week. A week later, we found the gut microbiota of KO mice gavaged with fecal microbiota of WT mice clustered differently from KO mice gavaged with sterile PBS (Fig. 4b). In phylum level, we found the relative abundance of Deferribacteres was increased in KO mice gavaged with fecal microbiota of WT mice (Fig. 4c-d). And in species level, we found Mucispirillum, which is a genus in the phylum Deferribacteres, was increased in KO mice treated with fecal microbiota of WT mice (Fig. 4e). Also, fecal microbiota transplantation ameliorated the decreased relative abundance of Prevotellaceae_UCG-001 and the increased relative abundance of Lactobacillales in KO mice ( Fig. 4fg).
Then functionally, we found, after being gavaged with fecal microbiota from WT mice, KO mice showed decreased self-grooming (Fig. 4h) and increased social behavior ( Fig. 4i-l). These results indicated that gut microbial dysbiosis was responsible for autism-like behavior in mice with deletion of EphB6.
Gut microbiota-mediated vitamin B6 metabolism regulated social behavior in EphB6-

deficient mice
Considering the abnormal behaviors were probably because of the problem of brain, so we tried to figure out how gut microbiota affected brain and subsequently caused autism-like behavior in EphB6deficient mice.
First, we tried to find the key region of brain affected by dysregulated gut microbiota in mice with deletion of EphB6, which was responsible for autism-like behavior. Studies on ASD patients or mouse models show that hippocampus, cerebellum and mPFC have been implicated in ASD [29,30]. After being processed with three-chambered social approach task, we found the protein expression of c-Fos was significantly increased in mPFC of KO mice compared with WT mice (Additional file 1: Figure S3ac). ASD has been generally considered to be caused by an increased ratio of synaptic excitation and inhibition and ASD children exhibit elevations in resting state neuronal activity [31]. So, whether mPFC was modulated by gut microbiota in KO mice needed to be further investigated. Because mPFC tissue was too small for some experiments and we used PFC tissue of mice in our next study.
The first question we asked was whether the bacteria could modulate mPFC directly. Unfortunately, we did not detect bacterial DNA or any bacterial colonies in PFC tissues of WT or KO mice (Additional file 1: Figure S3d-e). Unexpectedly, we found metabolites of gut microbiota from KO mice also induced social deficits in C57BL/6J mice (Additional file 1: Figure S3f-k). Were there some substances had been affected by gut microbial dysbiosis that caused social deficits in KO mice?
To found the metabolites that had been significantly changed, we detected metabolites in target tissue, that is PFC of KO mice, using non-targeted metabolomics strategies. Surprisingly, the metabolites in PFC were significantly different between the two groups of mice using orthogonal partial least squares discriminant analysis (Fig. 5a). KEGG pathway analysis showed 4 pathways that were significantly enriched in the differentially changed metabolites, including vitamin B6 metabolism pathway because of decreased relative abundance of pyridoxamine (PM) and pyridoxal 5'-phosphate (PLP) in PFC of KO mice (Fig. 5b-d).
Vitamin B6 in body is mainly from diet and gut bacteria's synthesis and then is absorbed in intestine.
Then we detected the level of vitamin B6 in feces, blood and PFC of mice. The increased level of pyridoxine (PN) in feces, decreased level of PM and PLP in plasma, and decreased level of PLP in PFC were found in EphB6-deficient mice ( Fig. 5e-j). A week after being gavaged with fecal microbiota from WT mice, KO mice had decreased level of PN in feces, increased level of PM and PLP in plasma and increased level of PLP in PFC compared with KO mice gavaged with sterile PBS (Fig. 5e-j). These results indicated that gut microbiota regulated the level of vitamin B6 in feces, blood and PFC of mice.
Then we wondered if vitamin B6 supplementation could ameliorate the autism-like behavior of KO mice. However, the intragastric supplementation of vitamin B6 did not ameliorate social deficits of KO mice (Additional file 1: Figure S4a-b). While one hour after being injected with 1 mg PLP intraperitoneally, KO mice had increased level of PLP in plasma (Fig. 5l), and increased social behavior were not changed in KO mice after the injection of PLP. Also, the injection of 1 mg or 2 mg PLP intraperitoneally had no effect on social behavior of C57BL/6J mice (Additional file 1: Figure S4c-e).
Moreover, after being fed without vitamin B6 for two weeks, C57BL/6J mice had decreased level of PLP in plasma and decreased social behavior ( Fig. 5r-u). Conclusively, our results hinted that there was a relation between gut microbiota-mediated defects of vitamin B6 metabolism and social deficits in EphB6-deficient mice.

Gut microbiota-mediated vitamin B6 metabolism regulated dopamine in PFC in EphB6deficient mice
Then we tried to clarify how the decreased vitamin B6 induced social deficits in mice. Vitamin B6, as a co-factor, has been implicated in more than 140 biochemical reactions in cells, including biosynthesis and catabolism of amino acid and neurotransmitters [32]. As the most important active substances in brain, we first detected the neurotransmitters in PFC of mice by high performance liquid chromatography (HPLC) and found similar levels of glutamate, GABA, glycine, aspartic acid, serine and glutamine, among WT and KO mice gavaged with sterile PBS or fecal microbiota from WT mice ( Fig. 6a). Interestingly, we found a decreased dopamine and an increased 5-HT in PFC of KO mice compared with WT mice (Fig. 6b). When treated with fecal microbiota from WT mice, KO mice had an increase in level of dopamine, but similar level of 5-HT, in PFC compared with KO mice gavaged with sterile PBS. While the level of noradrenaline, epinephrine and DOPAC were similar among the three groups of mice. More excitingly, after being gavaged with fecal microbiota from KO mice, the level of dopamine in PFC of SPF C57BL/6J mice had a decrease compared with C57BL/6J mice gavaged with fecal microbiota from WT mice (Additional file 1: Figure S5a To answer whether decreased dopamine contributed to autism-like behavior in EphB6-deficient mice and considering the fast metabolism of dopamine in brain, we injected the agonists of dopamine receptors into mPFC of mice. While deletion of EphB6 had no effect on mRNA expressions of dopamine receptors or tyrosine hydroxylase (Th) in mPFC or ventral tegmental area (VTA) (Fig. 6e).
As dopamine D1 receptor (D1R) had the highest expression in mPFC and then followed by dopamine D2 receptor (D2R) (Fig. 6f), we injected the D1R agonist (SKF38393) or D2R agonist (quinpirole) into mPFC of mice. We found KO mice had increased social behavior (Fig. 6g-j) after being injected with SKF38393 compared with KO mice injected with artificial cerebrospinal fluid (ACSF). While there were not any differences in C57BL/6J mice injected with ACSF or SKF38393 (Additional file 1: Figure S5d-f).
Differently, quinpirole did not increase social behavior in KO mice (Additional file 1: Figure S5g-i).
What's more, D1R antagonist induced a decreased social behavior in C57BL/6J mice ( Fig. 6k-n). In short, these results proposed dysregulated gut microbiota and vitamin B6 metabolism led to autismlike behavior by D1Rs-mediated pathway in EphB6-deficient mice.

Gut microbiota regulated E/I balance in mPFC of EphB6-deficient mice
D1Rs are generally considered to modulate GABAergic inhibition in PFC [33]. Also, imbalance between excitation and inhibition (E/I) in synaptic transmission and neural circuits has been implicated in ASD [34][35][36]. Moreover, the correction of E/I imbalance can normalize key autistic phenotypes in animal models of ASD [37].
Then to further investigate the cellular mechanism underlying gut microbiota-mediated autism-like behavior in EphB6-deficient mice, we recorded spontaneous excitatory postsynaptic currents (sEPSCs) and spontaneous inhibitory postsynaptic currents (sIPSCs) of mPFC pyramidal neurons, in WT and KO mice treated with sterile PBS or fecal microbiota from WT mice. The amplitude and frequency of sEPSCs were similar among the groups (Fig. 7a-e). The amplitude of sIPSCs was also similar among the groups, while the frequency of sIPSCs was decreased in KO mice and was rescued by transplantation of fecal microbiota from WT mice ( Fig. 7f-j). Additionally, we found a decreased frequency of sIPSCs in pyramidal neurons of mPFC in C57BL/6J mice gavaged with fecal microbiota of KO mice (Additional file 1: Figure S6a-h). D1R agonist, at a concentration of 10 μM, increased the frequency of sIPSCs in pyramidal neurons of mPFC in KO mice ( Fig. 7k-n), while the same concentration of D1R agonist had no effect on sEPSCs or sIPSCs recorded in pyramidal neurons of mPFC in WT mice (Additional file 1: Figure S6i-l). Collectively, these results indicated that gut microbiota modulated E/I balance, which was possibly regulated by dopamine, in pyramidal neurons of mPFC in EphB6-deficient mice.

Discussion
Increasing evidences, especially the clinical studies in ASD patients, suggest a functional link between gut microbiota and development of ASD. Besides, genomic and transcriptome studies for ASD patients discover many candidate genes for ASD. However, the functions of ASD-associated genes and the mechanisms linking gut microbiota and brain dysfunctions (gut-brain axis) in ASD are still not well-established.
First in our study, we uncover EphB6 is an ASD-associated gene functionally. EPHB6 has been suggested as a candidate gene for ASD for a long time [15][16][17] and is found to be down-regulated in ASD patients [19,20]. Here, using our transgenic mouse models, we found deletion of EphB6 induced autism-like behavior in mice that mimicked the core symptoms of ASD patients fairly well. Using whole-genome sequencing and transcriptome analysis, researchers have found more than 1000 genes that are associated with ASD, including EPHB6 [15][16][17][18], EPHA1 [15], EPHB2 [38]. Our study uncovers the functional role of EphB6 in ASD and suggests EphB6-deficient mice can be used as a new mouse model of ASD.
Secondly, we find gut microbial dysbiosis is required for autism-like behavior in EphB6-deficient mice.
Most ASD patients have serious GI problems [5,6,26] and changed composition of gut microbiota [39,40]. Moreover, microbiota transfer therapy can improve GI and autistic symptoms in ASD children [11,12]. Our study also suggests the probable role of gut microbiota on treating core symptoms of adult ASD patients. Eph families have the important role in regulating epithelial homeostasis by the interaction with epithelial cell adhesion and junction proteins [22]. Cldn4 can interact with EphA2 and ephrin-B1 to affect the tight junction integration [41,42]. AF-6 can be recruited to cell-cell contacts in MDCK and 293T cells by interacting with Eph receptors, including EphB6 [43,44]. The ablation of EphB6 may induce the dysregulated interaction between Eph families and junction proteins that leads to increased intestinal mucosal permeability and then gut microbial dysbiosis in mice.
Thirdly, we find dysfunction of vitamin B6 metabolism is key for gut microbiota-mediated autism-like behavior in EphB6-deficient mice. We found the decreased level of vitamin B6 in plasma and PFC of EphB6-deficient mice was rescued by transplantation of fecal microbiota from wild-type mice.
Moreover, injection of vitamin B6 intraperitoneally rescued social deficits of EphB6-deficient mice.
More interestingly, PLP has been found to have an unbelievable low level in ASD children compared with controls [45]. Date to 1960s, a lot of clinical studies have used vitamin B6 to treat ASD children, and most studies have reported the improved autistic symptoms in ASD children [46][47][48]. While there are also reports of the non-effect of vitamin B6 on ASD patients [49]. Considering the complicated causes of ASD, we think vitamin B6 is effective for a part of ASD patients, such as ASD patients with down-regulated expression of EPHB6. Vitamin B6 cannot be synthesized by the body itself and the source of vitamin B6 in body is mainly from diet and bacteria's synthesis via intestinal absorption. So, normal intestinal functions are important for the homeostasis of vitamin B6 in body. The intestinal absorption of vitamin B6 is pH dependent with higher uptake at acidic compared with alkaline pHs [50]. The more alkaline environment in gut of EphB6-deficient mice (Additional file 1: Figure S1i) may cause decreased absorption of vitamin B6. How the changed bacteria affect the gut pH and vitamin B6 level in feces and blood need more exploration. Overall, our study finds a new modulated role of gut microbiota on vitamin B6 and proves the ameliorative role of gut microbiota-mediated vitamin B6 on social deficits in EphB6-deficient mice.
Finally, we functionally establish the mechanisms linking gut microbiota and brain dysfunctions (gutbrain axis) in EphB6-deficient mice. Gut-brain axis has been generally considered to be involved in psychiatric diseases. However, there are few studies on how brain is specially regulated by gut microbiota. In our study, we found dopamine in PFC of EphB6-deficient mice was regulated by gut microbiota-mediated vitamin B6. In ASD patients, there are lower medial prefrontal dopaminergic activity [51]. After being given vitamin B6, autistic children have a reduced urinary homovanillic acid, which suggests an improved dopamine metabolism [52]. These studies suggest the regulated role of vitamin B6 in dopaminergic metabolites in ASD patients. Previously, Sgritta reported the modulated VTA plasticity by Lactobacillus reuteri in ASD mouse models [10]. Here, our study showed a new regulatory role of gut microbiota on dopamine in PFC by modulating vitamin B6 in EphB6-deficient mice. What's more, we found the ameliorative role of D1R agonists in social behavior and the modulated E/I balance in mPFC by gut microbiota in EphB6-deficient mice. Activating D1Rs in PFC can increase frequency of sIPSCs of pyramidal neurons, while D2R agonist does not have the same effect [33]. The modulation of social behavior by D1Rs was probably because of its modulation of GABAergic inhibition in EphB6-deficient mice. Collectively, our study indicates decreased dopamine is induced by dysregulated gut microbiota-mediated defect of vitamin B6 and then contributes to E/I imbalance and social deficits in EphB6-deficient mice.

Conclusions
In summary, our study uncovers a key role for gut microbiota in autism-like behavior of EphB6deficient mice. Mechanistically, gut microbiota-mediated defect of vitamin B6 metabolism regulates autism-like social behavior by decreasing dopamine levels and inducing E/I imbalance in mPFC in EphB6-deficient mice. Our study suggests a new ASD mouse model, proves the important role of gut microbiota in genetic factor-induced autism and provides a new insight into gut-brain-microbiota axis.

Fecal microbiota transplantation
Fresh feces of healthy male EphB6 +/+ and EphB6 -/mice (8 mice for each group from at least 3 cages) were collected from the disinfected anus into new sterile tubes every day before the experiment to promise microbial vitality [4]. Then the fresh feces were weighted, mixed with sterile PBS at a dilution ratio of 1 mg/10 μL or 1 mg/20 μL and centrifuged at 900 x g for 3 min. The supernatant was collected and gavaged orally to each mouse (10 mL/kg) for 5 or 7 consecutive days. All the mice were handled aseptically.
For fecal microbiota or metabolite transplantation, after the fresh feces were weighted, mixed with sterile PBS at a dilution ratio of 1 mg/10 μL and centrifuged at 4000 x g for 10 min, the supernatant and precipitate were both collected. After being filtered by the filter with a pore size of 0.22 μm (Cat# SLGP033RS, Millipore, Darmstadt, Germany), the supernatant was orally gavaged to each mouse (10 mL/kg) for 7 consecutive days. After being resuspended in sterile PBS, centrifuged at 900 x g for 3 min, washed by sterile PBS twice, the precipitate was orally gavaged to each mouse (10 mL/kg) for 7 consecutive days.

Antibiotics treatment
Vancomycin (50 mg/kg, CAS: 123409-00-7, MP bio, California, USA), neomycin (100 mg/kg, CAS: 1405-10-3, MP bio) and metronidazole (100 mg/kg, CAS: 443-48-1, MCE, New Jersey, USA) were mixed using sterile drinking water [4]. Then the mixture was orally gavaged to 3-week-old or 6-week-old SPF C57BL/6J mice twice a day for 5 or 7 consecutive days and the amount of infusion was based on the weight of mice. The mixture was prepared every day and used freshly. During the treatment, ampicillin (1 mg/mL, CAS: 69-52-3, MP bio) was added into the drinking water of mice and changed with fresh solution every 3 days. For 3-week-old SPF C57BL/6J mice, the antibiotics treatment lasted for 5 days [53]. All the mice were handled aseptically.

Western blot analysis
After abdominally anesthetized with phenobarbital sodium (60 mg/kg), the brain of mouse was quickly removed, put into an ice-cold mouse brain mold (Cat# 68713, RWD, Shenzhen, China) and sliced.
Then posterior mPFC, hippocampus and cerebellum of mice were cut out. The total proteins of tissues were extracted using the lysis buffer (Cat# P0013B, Beyotime, Shanghai, China) and boiled in protein loading buffer. Equal amounts of the denatured protein samples were electrophoresed in 6-10% polyacrylamide gel containing 0.1% SDS and transferred to polyvinylidene fluoride (PVDF) membranes with pore size of 0.45 μm (Cat# IPVH00010, Millipore). Then the PVDF membranes were incubated with primary antibodies at 4°C for at least 12 hr. After that, the samples were incubated with secondary antibodies for about 2 hr at room temperature (Cat# BA1050 and Cat# BA1054

Quantitative reverse transcription PCR (qRT-PCR)
After anesthetized with phenobarbital sodium (60 mg/kg), different tissues of mice were quickly removed and put into liquid nitrogen, including colon, colonic epithelium, spleen and lung. Then posterior mPFC and VTA were sectioned out using ice-cold mouse brain mold (Cat# 68713, RWD).
qRT-PCR was performed accordingly [54] by using a 7500 real-time PCR system (ABI, California, USA) and SYBR Premix Ex Taq (Cat# RR420A, Takara, Osaka, Japan). Normalized to the mRNA expression level of Gapdh or Actb, the mRNA expressions of other genes were evaluated using the method of ΔΔCt. All primers used in qRT-PCR were listed in Additional file 2: Table S1.

Hematoxylin-eosin staining
The staining of different tissues of mice with haematoxylin and eosin was performed accordingly [54].
Briefly, the tissues were immersed into 4% formaldehyde immediately for 24 hr. Then tissues were embedded in paraffin, sectioned, and stained with haematoxylin and eosin. The stained sections were observed using an optical microscope (Olympus, Tokyo, Japan).

Intestinal permeability assay
Mice were fasted for 4 hr before experiment, then FITC-dextran (50 mg/mL, Cat# 46944, Sigma Aldrich, Missouri, USA) was gavaged to mice (600 mg/kg) [8]. 4 hr after the oral gavage, the blood of mouse was collected by cardiac puncture. Then the blood was placed at room temperature for 1 hr before being centrifuged at a speed of 3000 rpm for 10 min. Then the supernatant was transferred to a new tube and centrifuged at a speed of 12000 rpm for 10 min at 4°C. The supernatant, which was the serum, was diluted with equal volume of PBS and 100 μL diluted serum was added to a 96-cell microplate. The concentration of FITC in serum was determined by Varioskan LUX microplate reader (Thermo Fisher Scientific, Massachusetts, USA) with an excitation of 485 nm and an emission wavelength of 528 nm. The serial diluted FITC-dextran (0, 0.5, 1, 2, 4, 6, 8, 10 μg/μL) was used as standards. Serum of mice administered with PBS was used as negative controls.

Vitamin B6-deficient mouse model
The formula of diet with normal vitamin B6 or without vitamin B6 was based on previous study [55].
Then 6-week-old SPF male C57BL/6J mice were fed with the diet with or without vitamin B6 for two weeks.

Behavioral studies
Mice used for experiments were male and naive. Mice were handled for 3 days before the experiments and habituated in the experiment room for at least 30 min before each test [56]. Mice were performed with different behavioral tests with a sequence or different mice were used for different behavioral tests which were mentioned in the figure legends. The sequence of different behavioral tests was self-grooming test, olfactory habituation/dishabituation test, three-chambered social approach task, marble burying test, open field test, social partition test, elevated plus maze and morris-water-maze test. Different behavioral tests were done with an interval of at least 2 days.
Self-grooming test was performed as previously described [57]. Generally, mouse was first placed in an empty crystal cage to habituate the cage for 10 min, then the time each mouse spent on selfgrooming was recorded during next 10 min by a double-blind experienced experimenter. Selfgrooming included face-wiping, scratching/rubbing of head and ears, and full-body grooming.
Between each trial, the apparatus was cleaned by 30% ethyl alcohol in water.
Marble burying test was performed as previously described [58]. Mouse was placed into an animal cage filled with fresh wood chip bedding with the depth of 5 cm. Regular pattern of glass marbles (5 rows of 4 marbles), which were placed 4 cm apart from each other, were regularly placed under the bedding and mouse was allowed to explore for 30 min. The number of buried (more than 50 percent of their depth in bedding) marbles was counted.
Social partition test was performed as previously described [59]. Mouse was individually housed in one side of the cage which was divided by a clear perforated partition with 0.6 cm-diameter holes, and the other side of cage was housed with a sex-and age-matched C57BL/6J mouse for 24 hr before experiment. At the first trial, the total time that the experimental mouse spent on sniffing partition with the familiar mouse on the other side during 5 min was recorded. Then the familiar mouse was replaced with a sex-and age-matched unfamiliar C57BL/6J mouse, the total time that the experimental mouse spent on sniffing partition with the unfamiliar mouse during 5 min was recorded.
In the last trial, unfamiliar mouse was instead replaced by familiar mouse and the time that the experimental mouse spent on sniffing partition as first trial was recorded. The time spent on sniffing partition in the three trials was recorded by a double-blind experienced experimenter.
Olfactory habituation/dishabituation test was performed as previously described [60]. Mouse was placed into a clean usual animal cage with thin bedding in a fresh room for 30 min before test. One swab saturated with water was given to mouse for 2 min, and then quickly replaced by another swab saturated with water for the following 2 min, then third swab saturated with water was given to mice for another 2 min quickly. Then other odors were given to mouse similarly. The sequence of given odors was water, almond extract, imitation banana flavor, odor of soiled bedding from mice and odor of soiled bedding from another cage of mice. Water, almond extract (dilution of 1:100) and imitation banana flavor (dilution of 1:100) were regarded as unsocial odors, while soiled bedding with the excrement of sex-and age-matched unfamiliar C57BL/6J mice were regarded as social odors. The time each mouse spent on sniffing the odorant swabs in every 2 min trial was recorded by a doubleblind experienced experimenter.
Three-chambered social approach task was performed as previously described [61]. The apparatus was divided into three rectangular clear chambers (60 cm x 40 cm x 22 cm) by two walls on which had two removable doorways (8 cm x 5 cm) that allowed mouse to access each chamber freely. After habituated to the middle chamber for 5 min, the mice were allowed to explore the three chambers freely for 10 min. For the sociability test, an age-and sex-matched C57BL/6J mouse was placed in the wire cage in one chamber while the wire cage in the other chamber was empty. Then the dividers were raised and the experimental mouse was allowed to freely explore all three chambers for 10 min.
For the social novelty test, another age-and sex-matched C57BL/6J mouse was placed in the empty wire cage described above. And the experimental mouse was originally placed in the center of the chamber and allowed to explore freely for 10 min after doorways were removed. Between each trial, In open field test, mouse was placed in the center of an open field chamber (40 cm × 40 cm × 30 cm) [62]. Exploratory behavior of mice was assessed by a session of 30 min and total distance was automatically recorded and analyzed by a VersaMax animal behavioral monitor system (Omnitech Electronics, Nova Scotia, Canada).
Elevated-plus-maze test was performed accordingly [62]. Briefly, mouse was put into the center of Morris water maze test was performed as before [63]. Generally, 4 trials were given to each mouse every day for 5 days. In each trial, the searching time for the mouse was no more than 1 min. A stay on the platform was 15 s. Intervals between each trial were no less than 1 min. On the sixth day, the probe test was performed by removing the platform and recording the swimming paths of mice in 1 min. The swimming paths of mice during the learning and test period were analyzed by EthoVison XT software (Noldus).

16S rRNA gene sequencing
Fecal samples of the experimental mice were collected and stored at −80°C before being performed.
Using QIAamp Fast DNA Stool Mini kit (Cat# 51604, QIAGEN, Venlo, Netherlands), genomic DNA of samples were extracted. The purity and concentration of the extracted DNA were detected using agarose gel electrophoresis. Bacterial DNA was amplified with the primers targeting V3-V4 regions (5'-TACGGRAGGCAGCAG-3', 5'-GGGTATCTAATCCT-3'). Then DNA was sequenced using MiSeq PE300 platform (Illumina, California, USA) by oe biotechnology company in shanghai. The raw data were treated and processed using QIIME software package (version 1.8.0). Then represent sequences of OTU were blasted in Silva database (version 123). The alpha diversity and beta diversity were analyzed using QIIME software package (version 1.8.0).

Metabolomic analysis
For untargeted metabolite analysis, PFC of mice were prepared and deproteinized with methanol.
Then the samples were analyzed using liquid chromatography-mass spectrometry by oe biotechnology company in shanghai. UPLC-Q-TOF/MS (ACQUITY UPLC I-Class, Waters, Massachusetts, USA) and ESI-QTOF/MS (Xevo G2-S Q-TOF, Waters) were used. The chromatographic column was the ACQUITY UPLC BEH C18 Column (1.7 µm, 2.1 mm X 100 mm, Waters). Mobile phase A was water contained with 0.1% formic acid and mobile phase B was acetonitrile contained with 0.1% formic acid.
For targeted metabolic analysis, PFC was pretreated with 0.4 M perchloric acid which contained 0.04% EDTA and 100 µL plasma was pretreated with 50 µL 5% trichloroacetic acid. 1 M NaOH was added to samples to quench acid.
For the analysis of amino acid neurotransmitters, high performance liquid chromatography (HPLC, Shimadzu, Kyoto, Japan) was used, with fluorescence detection system (Prominence RF-20A/20Axs, Shimadzu) and the C18 chromatographic column (Eclipse AAA, 4.6 x 150 mm, 5 μm, Agilent, California, USA). All the used reagents and liquid were chromatographically pure. Mobile phase A contained 20 mM sodium acetate solution (pH7.2), methyl alcohol and tetrahydrofuran which were at a volume ratio of 400:95:5. Mobile phase B contained 20 mM sodium acetate solution (pH7.2) and methyl alcohol which were at a volume ratio of 120:380. The gradient elution was 0%-63% mobile phase B in 0-10 min, 63% mobile phase B in 10-12 min, 63%-100% mobile phase B in 12-12. For the analysis of pyridoxal 5'-phosphate and pyridoxamine, TSQ Quantiva combined with Prelude SPLC System (Thermo Fisher Scientific) were used. All the used reagents and liquid were chromatographically pure. First, the separation of substances was performed using Prelude SPLC System with the C18 chromatographic column (Water Acquity UPLC HSS T3, 2.1 x 100 mm, 1.7 μm).
The flowing rate of mobile phase was 0.25 mL/min. Data were recorded using positive-ion electrospray ionization and the selected reaction monitoring mode. For pyridoxal 5'-phosphate, precursor ion was m/z 248.03, product ion was m/z 150.071 and collision energy was 16.067 V. For pyridoxamine, precursor ion was m/z 169.152, product ion was m/z 152.111 and collision energy was Deletion of EphB6 led to autism-like behavior and gut microbial disturbance in mice Transplantation of fecal microbiota from wild-type mice ameliorated autism-like behavior in adult EphB6-deficient mice Gut microbiota-mediated vitamin B6 metabolism regulated social behavior in EphB6deficient mice Figure 6 The modulated dopamine by gut microbiota-mediated vitamin B6 regulated social behavior of EphB6-deficient mice