We observed that the abundance of Faecalibacterium reduced in ASD individuals compared to typically developing (TD) individuals in several cohorts (Kang et al., 2018; Zhao et al., 2023; Pang et al., 2023; De Angelis et al., 2013; Ding et al., 2020), suggesting that Faecalibacterium may play a protective role in the context of ASD (Figure 1a and Supplementary Figure 1a). To obtain the microbial strains belonging to Faecalibacterium, we isolated and cultured the gut microbes derived from the fecal samples of TD and ASD individuals. We harvested 183 bacterial strains belonging to 65 species (Figure 1b). Focusing on Faecalibacterium, two Faecalibacterium prausnitzii strains and four Faecalibacterium hominis strains were obtained (Figure 1b). Notably, F. hominis was identified as a new species, and all strains of this species were isolated from the TD sample (Figure 1c). Consequently, we selected the strain F. hominis 4P-15 (4P-15) for further functional research (Figure 1b, the red triangle).

To explore whether 4P-15 could play a protective role in ASD by alleviating the ASD-like phenotypes in ASD animal model, we gavaged weaned BTBR mice daily for 4 weeks (Figure 2a). Subsequently, we evaluated ASD-relevant behaviors using the three-chamber test for social interaction deficits and the self-grooming test for stereotyped behavior. The results indicated that 4P-15 restored the stereotyped behavior in the self-grooming test (Figure 2b) and the impaired social interaction in the social ability session of the three-chamber test (Figure 2c and Supplementary Figure 2a), the core deficits of ASD. Furthermore, we conducted the open-field test for anxiety and the sucrose preference test for depression to assess the impact of 4P-15 on these behaviors. We did not observe any significant improvement of 4P-15 on anxiety or depression (Supplementary Figures 2b, c). Furthermore, we analyzed the correlation between the abundance of F. hominis in the feces and behavioral outcomes, and found a significant negative correlation between F. hominis abundance and stereotypic behavior in the F. hominis-gavaged BTBR mice (Supplementary Figures 2d, e).

Next, we explored the alterations associated with the mechanisms of the gut-brain axis in both the intestine and the brain in response to 4P-15 administration (Yu, 2021). Given that the genome of 4P-15 contained a microbial anti-inflammatory molecule showing 96.56% similarity (https://blast.ncbi.nlm.nih.gov/Blast.cgi) to F. prausnitzii (Auger et al., 2022) (Figure 3a), we hypothesized the anti-inflammatory properties of 4P-15 and the anti-inflammatory effect might mediate the protective effect of 4P-15 on ASD-like phenotypes. We analyzed six inflammatory cytokines in the cerebral cortex following gavage and found that 4P-15 did not affect the levels of these cytokines (Figure 3b). Then, we examined bacteria-derived compounds such as SCFAs and neurotransmitters that were previously reported to have altered levels in BTBR mice and participate in the modulation of CNS activity by gut microbes (Liu et al., 2024; Bove et al., 2024). Our results showed no significant changes in the levels of SCFAs in the feces or neurotransmitters in the brain after 4P-15 treatment (Supplementary Figures 2f, g).

Next, we explore the functional mechanisms via the gut-brain axis following the 4P-15 intervention, focusing on the intrinsic defects of the ASD mouse model we utilized. Previous studies reported excessive activation of the AhR signaling pathway in BTBR mice, characterized by increased expression of AhR (Uddin et al., 2023). Hyperactivation of AhR may lead to neurodevelopmental toxicity, contributing to the pathogenesis of ASD (Kimura et al., 2017; Siddiqui et al., 2021; Martin et al., 2022). In our investigation, we sought to determine whether 4P-15 could alleviate the hyperactivation of AhR by reducing the levels of AhR ligands, such as indole and its derivatives, which are primarily produced by gut bacteria and function mainly through AhR activation. We analyzed levels of indole and six indole derivatives in fecal samples (Figure 4a). Our results revealed elevated levels of indole and four derivatives, indole-3-acetic acid (IAA), indole-3-propionic acid (IPA), indole-3-lactic acid (ILA) and indole-3-aldehyde (IAId) in the feces of BTBR mice compared to C57BL/6J mice (Figure 4b). Notably, administration of 4P-15 led to a reduction in the levels of indole, IPA, and IAA (Figure 4b).

Subsequently, we delved into understanding how the administration of 4P-15 reduces the levels of intestinal indole and its derivatives. We first examined whether the genome of 4P-15 contained the enzymes capable of indole biotransformation and did not identify any functional elements (Ma et al., 2018) (Supplementary Figure 3a). We then assessed whether 4P-15 had the potential to decrease the concentrations of indole, indole-3-acetic acid (IAA), and indole-3-propionic acid (IPA) in vitro. By comparing the concentrations of these three compounds in the fermentation broth samples of 4P-15 compared with its YCFA medium, we observed no significant difference in IAA levels (Supplementary Figure 3b). Since the concentrations of indole and IPA were below the detection threshold in YCFA medium, we supplemented the YCFA medium with indole and IPA, respectively. Our findings indicated that 4P-15 failed to reduce the concentrations of both indole and IPA in the supplemented YCFA medium (Supplementary Figures 3c, d).

Next, our focus shifted to the influence of 4P-15 on the gut microbiota, particularly the abundance of indole-producing bacteria. Utilizing 16S rRNA gene sequencing of cecal contents, we found that 4P-15 administration did not significantly alter either alpha or beta diversity (Figure 4c and Supplementary Figures 3e–h). Subsequent analysis revealed a significant reduction in the abundance of two genera following 4P-15 treatment (Figure 4d). Notably, one of these genera, Peptococcus, is identified as a potential indole-producing bacterium (Lanigan, 1976; Schwan, 1979). These results suggest that 4P-15 might decrease intestinal indole concentrations by reducing the abundance of potential indole-producing bacteria.

We next examined whether 4P-15 gavage-induced reductions in intestinal indole and its derivatives influenced brain metabolite concentrations. 4P-15 administration significantly decreased IPA levels and modestly reduced indole concentrations in the brain (Figure 5a). To assess the potential neuroprotective effects of these metabolic changes, we prioritized AhR-regulated pathways directly relevant to ASD pathophysiology. While AhR modulates numerous downstream targets, we focused on glutamate transporters (Slc1a1, Slc1a2, and Slc1a3) and GABA receptors (Gabra1, Gabrb2, Gabrg2, and Gabbr1), given their dual rationale: (1) prior studies demonstrate AhR suppresses glutamate transporter expression (Silva-Parra et al., 2024) and GABA receptor transcription (Dever et al., 2016), and (2) ASD pathogenesis involves excitatory/inhibitory imbalance driven by glutamate-GABA signaling dysregulation. While 4P-15 administration did not significantly change Ahr mRNA level (Figure 5b and Supplementary Figure 4a), it notably upregulated the expression of glutamate transporters Slc1a2 and GABA receptors Gabra1, Gabrb2, and Gabbr1 (Figures 5c, d)

To investigate how AhR ligand dynamics influence its downstream targets, specifically glutamate transporters and GABA receptors, we analyzed correlations between brain indole/IPA levels and AhR pathway gene expression in BTBR mice gavaged with 4P-15. Notably, IPA concentrations exhibited significant inverse associations with Slc1a2/Slc1a3 expression, while indole levels showed a significant negative correlation specifically with Slc1a2 (Figure 5e and Supplementary Figure 4b). Together, these data suggest that 4P-15 derepresses glutamate transporters and GABA receptors through diminished AhR signaling by reducing indole-related AhR ligands, thereby alleviating excitatory-inhibitory imbalance.

The fecal samples from the typically developing (TD) individual were obtained as previously described (Liu et al., 2021). The fecal samples from the ASD individual used for gut microbial isolation were collected with approval from the Ethics Committee of Peking Union Medical College Hospital under the ethical approval number ZS-1393. The samples intended for bacterial isolation were kept fresh and transported to an anaerobic workstation (AW500, Electrotek, UK) within 2 hours of collection. The samples were suspended in anaerobic PBS and filtered through a 40 μm cell strainer. The resulting suspensions were divided into two portions: one portion was directly cultured on modified YCFA agar plates (YCFA agar medium supplemented with glucose at 1 g/L, hemin at 0.01 g/L, and L-cysteine at 1 g/L) and BHI agar plates supplemented with 5% defibrinated sheep blood. The other portion underwent alcohol pretreatment before being cultured on a different set of modified YCFA agar plates (YCFA agar medium supplemented with glucose at 1 g/L, hemin at 0.01 g/L, L-cysteine at 1 g/L, and sodium taurocholate at 0.1%) (Browne et al., 2016).

Colony isolation and identification were carried out as described in our previous study (Liu et al., 2021). All colonies on the plates were picked after cultivation periods of 24 h, 3 days, and 7 days, followed by incubation at 37°C under anaerobic conditions in the corresponding media for a duration of 1–7 days. Subsequently, 500 μl of the culture media were collected, and the bacterial pellets were harvested for PCR-based amplification of the 16S rRNA gene sequences using primers 27F and 1492R. The sequences of the PCR products were identified by aligning them against the NCBI 16S rRNA sequence database and the EZBioCloud database (updated on August 30, 2024). The threshold for 16S rRNA gene sequence identity for novel species was set at 98.7% (Liu et al., 2021). The phylogenetic tree of isolated gut microbiome strains was constructed by MEGA (Version 11.0.13) using the Neighbour-Joining Tree method. The 16S rRNA gene sequences of the isolated bacteria are provided in Supplementary Table 1. Isolated bacteria from the TD sample have been preserved in hGMB (Liu et al., 2021).

All animal experiments complied with the National Institute of Health Guide for the Care and Use of Laboratory Animals. The permission for animal experiment procedures was granted by the Animal Ethics Committee at the Institute of Microbiology, Chinese Academy of Sciences.

Mice were maintained on a 12 h/12 h dark/light cycle (lights on at 7:00 am) in the animal facility at the Institute of Microbiology, Chinese Academy of Sciences, under specific pathogen-free conditions. Male and female BTBR T⁺ Itpr3^tf/J (BTBR) mice purchased from Jackson Lab (America) were crossed to obtain BTBR mice offspring. After weaning, BTBR mice were reared separately by gender. C57BL/6J mice were purchased from SPF (Beijing) Biotechnology Co., Ltd (Beijing, China). After weaning, the mice were group-housed in cages containing two to five mice per cage. Mice from different experimental groups (C57BL/6J mice gavaged with PBS, BTBR mice gavaged with PBS, and BTBR mice gavaged with 4P-15) were housed separately. Male mice and samples from male mice were used in this study.

Animal numbers: The number of animals used in the behavioral tests (Figures 2b, c and Supplementary Figures 2a–c) were 10 for C57BL/6J mice gavaged with PBS, BTBR mice gavaged with PBS, and BTBR mice gavaged with 4P-15. The number of animals used for the cytokine analysis (Figure 3b) were 10, 8, and 10 for C57BL/6J mice gavaged with PBS, BTBR mice gavaged with PBS, and BTBR mice gavaged with 4P-15, respectively. The number of animals used in the 16S rRNA sequencing (Figures 4c, d and Supplementary Figures 3e–h) were 10 for C57BL/6J mice gavaged with PBS, BTBR mice gavaged with PBS, and BTBR mice gavaged with 4P-15. The number of animals used for intestinal indole detection (Figure 4b) were 10, 20, and 16 for C57BL/6J mice gavaged with PBS, BTBR mice gavaged with PBS, and BTBR mice gavaged with 4P-15, respectively. The number of animals used for the expression analysis of AhR, glutamate transporters and GABA receptors (Figures 5b–d) were 10 for C57BL/6J mice gavaged with PBS, BTBR mice gavaged with PBS, and BTBR mice gavaged with 4P-15. The number of animals used for neurotransmitter analysis (Supplementary Figure 2f) were 10 for BTBR mice gavaged with PBS and BTBR mice gavaged with 4P-15. The number of animals used for SCFA analysis (Supplementary Figure 2g) were 9 and 10 for BTBR mice gavaged with PBS and BTBR mice gavaged with 4P-15, respectively.

F. hominis 4P-15 was grown anaerobically in a modified YCFA medium supplemented with 3 mM acetate (1M acetate-sodium acetate buffer) to stimulate bacterial growth (Duncan et al., 2002). The dosage of F. hominis 4P-15 chosen for colonization was based on previous publications (Yu et al., 2022b; Martín et al., 2015; Miquel et al., 2015), which reported effective dosages ranging from 1 × 10⁸ to 1 × 10⁹ colony-forming units (CFU). For each mouse in the F. hominis 4P-15 administration group, 4 × 10⁸ CFU were suspended in 200 μl PBS and delivered by gavage immediately. For the control group, an equal volume of PBS was given. The wean C57BL/6J or BTBR mice at 4 weeks old were gavaged with PBS or 4P-15 for a duration of 4 weeks, after which fecal samples were collected. The mice underwent behavioral tests at 8 weeks of age or were sacrificed for brain tissue collection. The 11-week-old mice that completed behavioral tests were sacrificed. The gavage regimen began when the mice were 4 weeks old and continued until they were sacrificed. The numbers of mice for gavage were 10, 20, and 20 for C57BL/6J mice gavaged with PBS, BTBR mice gavaged with PBS, and BTBR mice gavaged with 4P-15, respectively.

The test mice were acclimated by handling them for three consecutive days before the experiments. Behavioral tests were carried out in the behavioral test room between 8:00 a.m. and 7:00 p.m., with the mice being transferred at least an hour in advance for acclimation (Yu et al., 2022b).

For testing the SCFAs, including acetic acid, propionic acid, butyric acid, pentanoic acid, and valeric acid in feces, 0.1 g feces of each sample were freeze-dried and extracted in 1 ml methanol by 10-min ultrasound bathing in iced water. After the centrifuge, the supernatant liquid was prepared for GC-MS analysis. GC-MS analysis was performed on a GCMS-QP2010 Ultra with an autosampler (SHIMADZU) and the DB-wax capillary column (30 m, 0.25 mm i.d., 0.25 μm film thickness; SHIMADZU). The temperature of the oven was programmed from 80°C to 140°C at a 20°C/min gradient, with a 1 min hold; to 290°C at 3.5°C/min, with a 15 min hold. Injection of a 1 μl sample was performed at 280°C. The carrier gas, helium, flowed at 1.2 ml/min. The electronic impact was recorded at 70 eV. Referenced to standard curves, the products from feces were calculated by peak area (Li et al., 2022).

Genomic DNA was extracted from mouse cecal contents or human feces by QIAamp PowerFecal^® Pro DNA kit (Qiagen, United States). The V3-V4 region of the 16S rRNA gene was amplified using the primer set 338F (5′-CCTACGGGNGGCWGCAG-3′) and 806R (5′-GACTACHVGGGTWTCTAAT-3′). PCR reactions were amplified using PremixTaq (TaKaRa, China), DNA products were purified by gel electrophoresis, and a sequencing adapter was connected to build the library followed by the standard protocol of NEBNext^®UltraTM DNALibraryPrepKitforIllumina^®. The 16S rRNA gene amplicon sequencing was conducted on Illumina novaseq 6000 PE250 platform.

The raw sequencing data of 16S rRNA gene sequences were processed using QIIME2 (v2021.4.0) (Bolyen et al., 2019). Sequences were trimmed to remove primer and adapter sequences. The trimmed raw reads were then analyzed with DADA2 (Callahan et al., 2016) to perform quality trimming, denoising, error-correction, paired-end read merging, chimera removal, and dereplication. The produced amplicon sequence variants (ASVs) were taxonomically classified according to the SILVA database (release 138) using “classify-sklearn,” after training the classifier on the S database (v.2022.10) (McDonald et al., 2024) with 341F/806R primers. Subsequently, the microbial profile table was exported for downstream analyses.

For diversity analysis, the microbial table was first normalized by total sum reads. Alpha diversity was assessed using the “diversity” function in the R package vegan, with statistical significance determined by the “aov” function. Then the “p.adjust” function (method = “bonferroni”) in R. Beta diversity analysis involved conducting principal component analysis (PCA) using the “prcomp” function (center = T, scale = T) in R and visualization with the ggbioplot R package. Multivariate analysis of variance in beta diversity analysis using distance matrices (PERMANOVA) was carried out with the “adonis2” function (permutations = 9999, method = “euclidean”) in the vegan package.

The metabolic gene content was estimated from 16S rRNA gene data using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2) pipeline with default parameters. ASVs with nearest-sequenced taxon index (NSTI) values >2 were filtered from the analysis. KEGG orthology (KO), and enzyme Commission number (EC number) metagenomes, and MetaCyc pathway abundances were predicted with PICRUSt2 (Renga et al., 2023).

Total RNA extracted from the left cerebral cortex of mice was used for cDNA preparation using a Hifair^® AdvanceFast One-step RT-gDNA Digestion SuperMix for qPCR (11151ES60, Yeasen, Shanghai, China). Real-time PCR analysis was performed using Hieff^® qPCR SYBR Green Master Mix (11203ES08, Yeasen, Shanghai, China) to determine the expression levels of glutamate transporters (Slc1a1, Slc1a2, Slc1a3), GABA receptors (Gabra1, Gabrb2, Gabrg2, Gabbr1), AhR (Ahr), cytochrome P450 (CYP) enzymes (Cyp1b1), and the Gapdh gene. The sequences of primers were provided in Supplementary Table 2.

False discovery rate (FDR)-adjusted P-values determined by two-sided Wilcoxon rank-sum tests was used for analyzing Feacalibacterium abundance differences in published cohorts (Figure 1a and Supplementary Figure 1). Mann-Whitney (Supplementary Figures 2b, c, f, g) and Student's t-tests (Figure 5a and Supplementary Figures 3b–d) were used for comparing data between two groups. One-way ANOVA with Tukey's multiple comparison tests [Figures 2b, c (right panel), Figures 3b, 5b–d] and one-way ANOVA with multiple comparisons by controlling FDR (Figures 4b, d) were used for multiple comparisons across three groups. Two-way ANOVA with Tukey's multiple comparison tests were used for comparing the effects of two factors on a response variable [Figure 2c (lefe panel) and Supplementary Figure 2a (left panel)]. Spearman correlation tests (Figure 5e and Supplementary Figure 4b) and Pearson correlation tests (Supplementary Figures 2d, e) in R were used to quantify associations. PERMANOVA with the adonis2 function (vegan package) using Euclidean distances was used for beta diversity analysis (Figure 4c).