Gut bacteria changes restore brain development and behavior through the gut-brain connection in autism-like zebrafish

All zebrafish experiments were approved by the Institutional Animal Care and Use Committee of Chongqing Medical University and performed in accordance with relevant guidelines to ensure animal welfare.

Katnal2^−/− zebrafish were obtained from Prof. Xiong Bo (Huazhong University of Science and Technology) and maintained at 28°C under a 14:10 light/dark cycle. Adult males and females were paired overnight at a 1:2 ratio in spawning tanks, and fertilized eggs were collected the following morning following light stimulation. Embryos were incubated at 28°C with daily water changes and removal of dead embryos; larvae were fed paramecia at 3 dpf and Artemia at 10 dpf (hatched under standard saline, aerated conditions at 28 ± 0.5°C for 48 h, with eggshells removed before feeding). GF zebrafish were generated from healthy embryos using our optimized sterile protocol (detailed in Text S1) [37]. Sterility was confirmed by culturing samples in TSB and BHI broth under oxic and anoxic conditions at 28°C for ≥72 h, and by plating on TSA and blood agar plates for ≥120 h. PCR assays were routinely performed to confirm the absence of bacterial contamination, using bacterial samples as positive controls.

Fecal matter for transplantation was collected from adult zebrafish after overnight housing in spawning tanks. The collection and processing protocol was adapted from established FMT methods [25, 38] before administration to GF recipients. Briefly, fresh fecal samples (100 mg) collected from adult AB and Katnal2 zebrafish were suspended in 10 ml phosphate-buffered saline (PBS), vortexed, and centrifuged at 500 g for 5 min. Then, the fresh supernatant was stored at 4°C for a moment before use, and long-term stored within 50% glycerol at −80°C. Fertilized eggs of AB and Katnal2 zebrafish were collected and divided into four groups (n = 100 each): AB control (CR-AB), Katnal2 control (CR-Katnal2), germ-free AB transplanted with Katnal2 feces (GF-AB+FMT1), and germ-free Katnal2 transplanted with AB feces (GF-Katnal2+FMT2). Control groups of CR-AB and CR-Katnal2 were maintained under standard conditions, while GF groups were generated according to an optimized sterile protocol (detailed in Text S1). At 5 dpf, GF-AB and GF-Katnal2 larvae had reached the mouth-opening stage with mature intestinal function, suitable for FMT and bacterial colonization studies [25, 39]. Larvae in the GF-AB and GF-Katnal2 groups, housed in 6-well plates, received 200 μL per well of the respective fecal supernatant, which was added to the culture medium during daily renewal. At 7 dpf, zebrafish from all experimental groups were sampled, flash-frozen in liquid nitrogen, and stored at −80°C for subsequent analysis. These samples were used for 16S rRNA genes sequencing of microbial composition, quantification of SCFAs, and RT-PCR analysis of gene expression.

After activation,A. muciniphilawas cultured anoxically in mucin-supplemented BHI broth (37°C, 24–72 h), and growth was monitored by OD600 to establish exponential phase kinetics. Colonies were Gram-stained for morphological assessment, and genomic DNA was extracted for 16S rRNA genes amplification (27F/1492R), sequencing, and phylogenetic confirmation against the NCBI database. Bacterial concentration was determined by correlating OD600 with colony-forming units: cultures in exponential phase were serially diluted (10⁻¹–10⁻⁷), plated on mucin-BHI agar, and incubated anoxically at 37°C, with CFU counts averaged across triplicate plates in three independent experiments. The bacterial concentration corresponding to OD600 absorbance was calculated using the following formula:

where C is the bacterial solution concentration, N is the average number of colonies on the plate at a given dilution, V is the volume of dilution used for plating (mL), and M is the dilution factor.

Fresh bacterial suspensions were labeled with CM-DiI dye (10 μg/mL in PBS, 0.5% DMSO) for 20 min at 37°C, followed by 15 min at 4°C, washed by repeated centrifugation (12 000 rpm, 1 min), and examined under a fluorescence stereomicroscope to confirm labeling. At 5 dpf, 20 larvae per well were transferred to 6-well plates, and fluorescence imaging was performed every 24 h to track intestinal colonization of A. muciniphila for 72 h. Retention and localization within the gut were quantified by fluorescence intensity.

The growth and developmental indices of zebrafish were recorded daily, including survival rate, hatching rate, body length, body weight, and heartbeats per 10 s. At 7 dpf, six larvae were randomly selected from each group for the FMT experiment (CR-AB, CR-Katnal2, GF-AB+FMT1, GF-Katnal2 + FMT2) and the A. muciniphila treatment (CR-AB, CR-AB+Akk, GF-AB, GF-AB+Akk), and placed individually in 24-well plates (one larva per well). After 30 min of acclimatization, measurements were taken simultaneously across all groups. Zebrafish movements were recorded using high-speed infrared cameras operating at 157 frames per second under two lighting conditions: complete darkness for 10 min and alternating light–dark cycles for 50 min. Video data were analyzed using EthoVision XT tracking software (Noldus, Wageningen, Netherlands).

The total genomic DNA was extracted from zebrafish using the E.Z.N.A. Soil DNA Kit. The concentration and purity of the extracted DNA were verified by 1% agarose gel electrophoresis. The V3-V4 hypervariable region of the 16S rRNA gene was amplified using barcoded primers 341F (5′-CCTAYGGGRBGCASCAG-3′) and 806R (5′-GGACTACNNGGGTATCTAAT-3′). All PCR reactions were carried out using Phusion High-Fidelity PCR Master Mix (New England Biolabs) under the following thermal cycling conditions: initial denaturation at 98°C for 1 min; 30 cycles of denaturation at 98°C for 10 s, annealing at 50°C for 30 s, and elongation at 72°C for 30 s; and a final extension at 72°C for 5 min. PCR products were detected by electrophoresis on a 2% agarose gel, then pooled and purified to construct sequencing libraries using the NEB Next Ultra DNA Library Prep Kit (Illumina, USA). Library quality was assessed on an Agilent 5400 system (Agilent Technologies, USA), and sequencing was performed on an Illumina platform to generate 250 bp paired-end reads. Bioinformatics analysis was performed by merging and processing raw FASTQ files using the QIIME 2 dada2 plugin to generate a feature table of amplicon sequence variants (ASVs). Differential abundance analysis was conducted using ANOVA, LEfSe, and DESeq2 to identify bacterial taxa with significant differences among groups. Alpha diversity indices were calculated at the feature level to estimate microbial diversity within individual samples, while principal coordinate analysis (PCoA) was used to visualize structural variation in microbial communities across samples. Spearman correlation analysis was employed to assess associations between bacterial taxa and key physiological indices. In addition, potential functional profiles of microbial communities were predicted using PICRUSt based on KEGG Ortholog (KO) annotations (Test S2).

Total RNA was extracted from zebrafish larvae using Trizol reagent, and purity was confirmed by spectrophotometry (A260/280 = 1.8–2.1; A260/230 > 1). RNA was reverse-transcribed into cDNA using a commercial reverse transcription kit, and RT-PCR quantified gene expression with SYBR Green (TB Green II) and zebrafish-specific primers (Table 1, designed via NCBI Primer). Cycling conditions were 95°C for 3 min, followed by 40 cycles of 95°C for 10 s, 60°C for 20 s, and 72°C for 30 s, with dissociation curve analysis to confirm specificity. All reactions were performed in triplicate, and relative expression was calculated using the 2^-ΔΔCT method with gapdh as the internal control.

Thirty zebrafish per sample were weighed, homogenized in PBS (1:9, w/v), and centrifuged at 3000 rpm for 10 min. Supernatants were stored at −80°C until analysis. The total SCFA level in the samples was quantified using the ELISA kit.

All data were analyzed using SPSS (version 21.0) with appropriate statistical tests, including one-way ANOVA, Student’s t-test, or the nonparametric Kruskal-Wallis test. Developmental indexes, behavioral data, and RT-qPCR were analyzed and visualized using Origin software (Version 2024b) and Graphpad Prism10 with ANOVA (Brown-Forsythe test, Bartlett test, Tukey multiple comparisons). All data were presented with the mean ± standard error (SEM) for each group with at least three replicates. The statistical symbols of ^*, ^**, and ^*** were indicated P < .05, .01, and .001, representing significant differences between control and treatment groups.

GF zebrafish were successfully established with no detectable bacterial growth. To evaluate microbiota effects, FMT was performed between AB and Katnal2 fish (Groups: CR-AB as AB control, CR-Katnal2 as Katnal2 control, GF-AB+FMT1 as germ-free AB fish transplanted with Katnal2 feces, and GF-Katnal2 + FMT2 as germ-free Katnal2 fish transplanted with AB feces), and developmental indices were compared with those of the CR controls (Fig. 1). The hatching rate was significantly reduced in CR-Katnal2 (11.1%) versus CR-AB (30.5%, P < .01), indicating impaired early development (Fig. 1A). Body length was decreased in GF-AB+FMT1 (4084.3 μm), CR-Katnal2 (3890.6 μm), and GF-Katnal2 + FMT2 (4044.0 μm) compared with CR-AB (4278.6 μm, all P < .001). However, both FMT groups showed partial rescue relative to CR-Katnal2 (P < .001) (Fig. 1B). Body weight showed a similar trend: CR-AB averaged 0.54 mg/30 fish, while GF-AB+FMT1 (0.44 mg, P < .01) and CR-Katnal2 (0.47 mg, P < .05) were lower; GF-Katnal2 + FMT2 (0.52 mg) exceeded both GF-AB+FMT1 (P < .01) and CR-Katnal2 (P < .05) (Fig. 1C). Heart rate was reduced in CR-Katnal2 (21.9 beats/10 s, P < .05) versus CR-AB (23.1 beats/10 s) and further decreased in GF-Katnal2 + FMT2 (20.9 beats/10 s, P < .001). In contrast, GF-AB+FMT1 showed a significant increase, compared to CR-Katnal2 (P < .05), while GF-Katnal2 + FMT2 also exhibited partial recovery (P < .05) (Fig. 1D).

Video tracking revealed distinct locomotor patterns among groups (Fig. 1). GF-AB+FMT1 zebrafish showed sparse, edge-restricted trajectories, compared to CR-AB, while CR-Katnal2 exhibited reduced movement; in contrast, GF-Katnal2 + FMT2 displayed denser trajectories and greater activity than GF-Katnal2 (Fig. 1E and F). Quantitatively, CR-Katnal2 showed significant reductions in distance traveled, velocity, movement duration, manic frequency/duration, and active frequency/duration relative to CR-AB (Fig. 1G–N). Both FMT groups partially rescued these deficits: GF-AB+FMT1 and GF-Katnal2 + FMT2 exhibited longer movement times, higher activity frequencies, and shorter static durations than CR-Katnal2, with GF-Katnal2 + FMT2 showing the most robust recovery (Fig. 1G–N). In the light–dark test, CR-AB, GF-AB+FMT1, and GF-Katnal2 + FMT2 responded more sensitively to stimulation, whereas CR-Katnal2 showed minimal activity. GF-Katnal2 + FMT2 activity approached CR-AB levels, indicating FMT’s therapeutic effect on ASD-like zebrafish (Fig. 1O).

To directly evaluate whether the microbiota regulation could improve the ASD-like zebrafish development to normal, the GF AB and Katnal2 larvae were subjected to FMT between the two lines (Fig. 2A). Gene expression analyses revealed significant alterations in pathways related to brain development and neurotransmission (5-HT, DA, GABA) (Fig. 2B and Fig. S1). Among developmental genes, spra was unchanged in GF-AB+FMT1, but dhfr was reduced (P < .05); GF-Katnal2 + FMT2 also showed lower dhfr, comparable to CR-Katnal-2 (Fig. S1A and B). pah was significantly lower in GF-AB+FMT1 than in both CR-AB and CR-Katnal2. For serotonergic markers, tph1a was downregulated (P < .01), and htr3a upregulated (P < .05) in GF-AB+FMT1 versus CR-AB, while GF-Katnal-2 + FMT2 showed no change (Fig. S1C–E). Dopaminergic th was suppressed in GF-AB+FMT1 and CR-Katnal2 compared with CR-AB (P < .05), but significantly upregulated in GF-Katnal2 + FMT2 versus CR-Katnal2 (P < .01) (Fig. S1F). For GABA-related genes, gad1b was reduced in GF-AB+FMT1 (P < .05), with no difference in GF-Katnal2 + FMT2 versus CR-Katnal2 (Fig. S1G). Carbonic anhydrases showed group-specific regulation: GF-AB+FMT1 decreased ca4a and ca14 (P < .05), while GF-Katnal2 + FMT2 elevated the expression of ca2, ca7, and ca14 (P < .05), with no change in ca4a (Fig. S1H–K). These findings indicated the distinct regulatory effects on neuro-related key genes of FMT from AB and ASD-like zebrafish donors, in which the neuro-related signaling pathways were inhibited, but the normal microbiota rescued Katnal2 fish (Fig. 2B).

Total content of SCFAs were quantified across experimental groups (Fig. 2C). No statistically significant differences were observed among the CR-AB (106.76 ng/mL), GF-AB+FMT1 (107.11 ng/mL), CR-Katnal2 (104.29 ng/mL), and GF-Katnal2 + FMT2 (110.34 ng/mL) groups, although a increasing trend was noted in the GF-Katnal2 + FMT2 group compared to CR-Katnal2. However, to indirectly assess SCFA levels, the ffar2 gene expression was analyzed. Compared with the CR-AB group, the GF-AB+FMT1, CR-Katnal-2, and GF-Katnal2 + FMT2 groups exhibited marked downregulation of ffar2 (Fig. 2C). GF-Katnal2 + FMT2 larvae showed significantly higher ffar2 expression than CR-Katnal2 zebrafish, indicating potential regulatory effects of transplantation with AB zebrafish fecal solution.

Under anoxic conditions at 37°C, A. muciniphila formed milky-white colonies on mucin-supplemented BHI agar within 24 h (Fig. 3A). DNA extracted from the strain produced a distinct 327 bp band on agarose gel electrophoresis (Fig. 3B), and Gram staining confirmed its oval, Gram-negative morphology (Fig. 3C). Growth monitoring by absorbance revealed entry into the exponential phase at ~10 h (Fig. 3D). CM-DiI staining produced granular red fluorescence (Fig. 3E), enabling in vivo tracking. In zebrafish, intestinal fluorescence increased at 24 h, declined by 48 h, and persisted through 72 h, with signals progressing from the gastrointestinal tract to the intestines and cloaca, confirming colonization and persistence of the probiotic (Fig. 3F and G).

Developmental indices were assessed in CR-AB, CR-AB+Akk, GF-AB, and GF-AB+Akk zebrafish to evaluate the effects of A. muciniphila intervention (Fig. 3). Hatching rates at 48 hpf and body weight did not differ significantly among groups, indicating that sterilization and A. muciniphila exerted minimal impact on these two parameters (Fig. S2A and B). Body length was significantly shorter in GF-AB (3628.8 μm) and GF-AB+Akk (3749.2 μm) than in CR-AB (4082.33 μm, P < .001), although GF-AB+Akk fish were longer than GF-AB (P < .05), indicating partial developmental recovery (Fig. 3H). Heart rate decreased in GF-AB (20.5 beats/10 s) compared with CR-AB (22.6 beats/10 s, P < .001), but was restored in GF-AB+Akk (22.5 beats/10 s), which was significantly higher than GF-AB (P < .001) and comparable to CR-AB+Akk (23.0 beats/10 s) (Fig. 3I). While, developmental indices of CR-Katnal2, CR-Katnal2 + Akk, GF-Katnal2, and GF-Katnal2 + Akk zebrafish showed the A. muciniphila effects on ASD-like animal models (Fig. 3). Hatching rates at 48 hpf and body weights were not differences among groups, indicating no influence of sterilization or A. muciniphila treatment on these two basic aspects (Fig. S2C and D). In contrast, body length was significantly reduced in GF-Katnal2 (3370.4 μm) and GF-Katnal2 + Akk (3533.6 μm) compared with CR-Katnal2 (3906.8 μm, P < .001). However, A. muciniphila supplementation partially rescued the growth of GF-Katnal2 + Akk fish, which appeared significantly longer than GF-Katnal-2 (P < .01) (Fig. 3J). GF Katnal2 mutants exhibited a significantly reduced heart rate (19.5 beats/10 s) compared to their CR counterparts (22.3 beats/10 s; P < .001). This deficit was significantly ameliorated by GF-Katnal2 + Akk, which restored the heart rate to 21.6 beats/10 s, a value comparable to the control baseline (Fig. 3K).

Behavioral performance of AB zebrafish and A. muciniphila-treated larvae was assessed at 7 dpf (Fig. 4). CR-AB+Akk fish displayed stable edge-oriented trajectories similar to CR-AB, whereas GF-AB and GF-AB+Akk exhibited weaker, disordered movements with reduced activity and prolonged dwell times; however, GF-AB+Akk showed greater mobility, higher activity intensity, and shorter static periods than GF-AB (Fig. 4A). Traveled distance was unchanged in CR-AB+Akk versus CR-AB, but significantly reduced in GF-AB and GF-AB+Akk, with GF-AB+Akk covering more distance than GF-AB (Fig. 4B). Velocity and travel time followed similar patterns: both were reduced in GF-AB and partially restored in GF-AB+Akk, while CR-AB+Akk remained comparable to CR-AB (Fig. 4B). Static duration increased in GF groups, with GF-AB+Akk trending lower than GF-AB. Hyperactivity measures were also impaired: manic frequency and duration were the lowest in GF-AB, but both were significantly improved in GF-AB+Akk (Fig. 4B). Likewise, activity frequency and duration were strongly reduced in GF-AB but significantly higher in GF-AB+Akk, indicating the positive effects of probiotic on fish with partial behavioral rescue (Fig. 4B). In light–dark stimulation experiments, all groups increased velocity due to light off and decreased under light on; CR-AB+Akk showed the highest light-induced activity, followed by CR-AB, GF-AB+Akk, and GF-AB (Fig. 4C).

In the behavioral analyses of A. muciniphila-treated Katnal2 larvae (Fig. 4D), CR-Katnal2 + Akk fish exhibited gentle changes in movement trajectories compared to CR- Katnal2, whereas GF-Katnal2 and GF-Katnal2 + Akk displayed reduced and scattered movement patterns characterized by chaotic trajectories, decreased activity levels, and prolonged stays at specific locations. GF- Katnal2 + Akk showed denser movement patterns, increased trajectory density, enhanced activity levels, and shorter static than GF-Katnal2 (Fig. 4D). Traveled distance and velocity in CR-Katnal2 + Akk were alike to CR-Katnal2 fish, but significantly reduced in GF-Katnal2 and GF-Katnal2 + Akk, with Akk treatment partial recovery (Fig. 4E). Similarly, the moved time was less in GF-AB and GF-Katnal2 + Akk than CR-Katnal2, though Akk supplement exceeded GF-Katnal2 (Fig. 4E). Static duration increased in both GF groups, that of supplemented with Akk bacteria group was significant shorter static duration than other groups (Fig. 4E). Hyperactivity measures were impaired in GF-Katnal2, but GF-Katnal2 + Akk showed significantly higher manic frequency and activity frequency/duration than GF-Katnal2, indicating the absence of microbiota adverse effects and probiotic-mediated behavioral rescue (Fig. 4E). In light–dark testing, all groups increased velocity under dark and decreased with light on; CR-Katnal2 + Akk showed the strongest response to light–dark stimulation, followed by CR-Katnal2, GF-Katnal2 + Akk, and GF-Katnal2 (Fig. 4F).

To discover the differences between ASD-like animal models and controls, composition of gut microbiota in AB and Katnal2 zebrafish larvae (7d) and adult (90d) was profiled by 16S rRNA genes (V3–V4) sequencing (Fig. 5). The α-diversity analysis (Shannon, Simpson, observed ASVs) revealed significantly greater diversity and abundance in adult AB zebrafish compared with Katnal2 (Table S1). The principal coordinate analysis (PCoA) clearly separated the two groups (Fig. S3A), and ASV analysis showed 504 features in AB and 209 in Katnal2, with 46 shared (Fig. S3B). At the phylum level, Proteobacteria dominated Katnal2 (87.67%) but were less abundant in AB (39.57%), whereas Firmicutes, Actinobacteria, Fusobacteriota, and Bacteroidota were enriched in AB (Fig. 5A and Fig. S3C). At the genus level, Stenotrophomonas was strikingly higher in Katnal2 (84.23% vs. 0.09% in AB), while Cetobacterium, Shewanella, Mycobacterium, Plesiomonas, and Vibrio were all more abundant in AB (Fig. 5A and Fig. S3D). LEfSe analysis (LDA > 2, P < .05) confirmed distinct microbial signatures: AB zebrafish were enriched in Mycobacterium, Shewanella, Vibrio, and Enterobacterales, whereas Katnal2 were dominated by Stenotrophomonas, Chryseobacterium, and Comamonas genera (Fig. 5B and Fig. S3E and F). These findings highlighted profound microbiota differences between AB and Katnal2 zebrafish, suggesting potential microbial contributions to ASD-related phenotypes.

At 7 dpf, α-diversity indices showed minimal differences across groups (Table S1), and PCoA revealed no clear separation between AB and Katnal2 larvae (Fig. S4A). Venn analysis identified 843 features in AB and 246 in Katnal2, with 97 shared (Fig. S4B). Taxonomic profiling of the top 20 bacterial taxa showed Proteobacteria as the dominant phylum in both groups (88.1% in AB, 95.4% in Katnal2), with Bacteroidota and Actinobacteria relatively enriched in Katnal2, while Firmicutes and Desulfobacterota were more abundant in AB (Fig. 5A and Fig. S4C). At the genus level, Stenotrophomonas, Methylobacterium, Desulfovibrio, Paracoccus, and Pseudomonas accounted for >70% of the microbiota in both groups (Fig. 5A and Fig. S4D). AB larvae were enriched in Stenotrophomonas (69.5%), Desulfovibrio (6.4%), and Pseudomonas (1.6%), whereas Katnal2 larvae showed higher levels of Methylobacterium (10.4%) and Paracoccus (2.8%). LEfSe analysis further highlighted Bosea and Variovorax as discriminant taxa in Katnal2 larvae (Fig. 5C and Fig. S4E and F).

DESeq2 analysis revealed 12 significantly upregulated taxa (e.g. Methylobacterium) and 5 downregulated taxa (e.g. Bifidobacterium) in the larval CR-AB vs. CR-Katnal2 comparison. Variovorax and Desulfovibrio exhibited the most pronounced changes in abundance (Fig. 5B). The significantly upregulated taxa (e.g. Stenotrophomonas; Chryseobacterium) and downregulated taxa (e.g. Nocardia; Luteolibacter) were discovered in the adult CR-AB vs. CR-Katnal2 groups (Fig. 5C). Additionally, KEGG functions of microbiota were predicted, and distinct functional shifts among treatment groups were revealed. For example, Amino acid and Carbohydrate metabolism were elevated in AB from 7d to 90d but not in the Katnal2 period. Replication and repair, and Translation showed increased trends in both fish lines, while Drug resistance: antimicrobial was opposite in two lines, suggesting altered core metabolic pathways (Fig. 5D).

16S rRNA genes sequencing revealed that A. muciniphila colonization did not alter overall α-diversity or PCoA clustering (Table S1, Fig. 6A) but shifted microbial composition. In AB fish, Venn analysis showed 850 vs. 151 features in CR-AB and CR-AB+Akk, with 90 shared (Fig. 6B). Among which, Firmicutes phylum, Enterococcus, and Aeromonas genera were enriched, while Proteobacteria phylum, Stenotrophomonas, and Desulfovibrio declined after A. muciniphila intervention (Fig. 6C and D). In Katnal2 larvae, 16S rRNA genes analysis detected 270 and 199 microbial features in the CR-Katnal2 and CR-Katnal2 + Akk groups, respectively, with 73 features shared between them (Fig. 6B). Probiotic treatment shifted the microbial composition, increasing the relative abundance of Firmicutes while decreasing that of Proteobacteria (Fig. 6C). At the genus level, this was characterized by an increase in Stenotrophomonas and decreases in Methylobacterium and Paracoccus (Fig. 6D). Further, Simper analysis proved the species that most contributed to the community structures differences in CR-AB vs. CR-AB+Akk, CR-Katnal2 vs. CR-Katnal2 + Akk, and CR-AB+Akk vs. CR-Katnal2 + Akk sets, which highlighted the dynamic alteration and roles of Desulfobacterota, Firmicutes_A, Firmicutes_D, and Verrucomicrobiota after probiotic treatment (Fig. 6E). Developmental comparisons further showed distinct stage-specific shifts in AB and Katnal2 zebrafish, with larvae dominated by Proteobacteria and taxa, such as Stenotrophomonas, Reyranella, and Pseudomonadaceae, while adults were enriched in Firmicutes, Fusobacteriota, and genera including Cetobacterium, Shewanella, Mycobacterium, Bifidobacterium, and Enterococcus (Fig. S5 and Fig. S6). Together, these results demonstrated that while diversity remained stable, A. muciniphila selectively reshaped microbial composition in a lineage- and stage-dependent manner.

Treatment with A. muciniphila enhanced activity in GF-AB larvae, suggesting modulation of brain development via the microbiota-gut-brain axis and neurotransmitter involving serotonin (5-HT), dopamine (DA), and gamma-aminobutyric acid (GABA) (Fig. 7). Expression of brain-development genes pah and dhfr was altered, compared with CR-AB, the CR-AB+Akk group showed no changes in spra, pah, or dhfr, whereas both GF-AB and GF-AB+Akk exhibited significantly reduced pah expression (P < .01, .001) (Fig. S7A–C), indicating limited effects of A. muciniphila on AB brain development. For the 5-HT system, tph1a was suppressed in GF-AB (P < .001) but upregulated in CR-AB+Akk (P < .05), while htr3a was increased in GF-AB (P < .05) and downregulated in GF-AB+Akk (Fig. S7D and E). Dopaminergic th was elevated in CR-AB+Akk versus CR-AB (P < .05) (Fig. S7F). GABA-related gad1b was significantly inhibited in both GF-AB and GF-AB+Akk (P < .001) (Fig. S7G). Carbonic anhydrases showed divergent regulation: CR-AB+Akk upregulated ca2, ca4a, ca7, and ca14 (all P < .05), whereas GF-AB reduced ca2 and ca14 (P < .01), and GF-AB+Akk further suppressed ca2 (P < .01) and ca14 (P < .001), with no significant differences between GF-AB and GF-AB+Akk (Fig. S7H–K). These results indicated that the probiotic promoted neuro-related signaling pathways in CR-AB fish, while the GF-AB fish showed significantly suppressed and gently improved by A. muciniphila intervention (Fig. 7B). Whole-body SCFAs content did not differ significantly among groups: CR-AB zebrafish (106.8 ng/mL), CR-AB+Akk (117.4 ng/mL), GF-AB (111.5 ng/mL), and GF-AB+Akk (116.2 ng/mL) (Fig. 7C). However, compared with the CR-AB group, the GF-AB group exhibited a significant decrease in ffar2 gene expression, whereas the GF-AB+Akk group increased the ffar2 expression relative to GF-AB (Fig. 7C).

In terms of expression of neurodevelopment- and neurotransmitter-related genes in Katnal2 zebrafish, A. muciniphila supplementation did not alter spra, pah (Fig. 7D, Fig. S8), or dhfr expression in CR-Katnal2 + AKK fish. However, GF-Katnal2 + Akk fish elevated spra expression (P < .01) and, relative to GF-Katnal2, higher spra (P < 0.001) and dhfr (P < .05) levels (Fig. S8A–C). The GF-Katnal2 group showed a lower level of tph1a (P < .01) compared with CR-Katnal2, whereas A. muciniphila intervention restored its expression (P < .001) (Fig. S8D). In contrast, htr3a was upregulated in GF-Katnal2 (P < .001) but suppressed by A. muciniphila (P < .01) (Fig. S8E). Expression of th was inhibited in both GF-Katnal2 (P < .01) and GF-Katnal2 + Akk (P < .05) than CR-Katnal2 (Fig. S8F). No changes were observed in gad1b (Fig. S7G). A. muciniphila supplementation increased carbonic anhydrase expression, including ca2 (P < .01), ca4a (P < .05), ca7 (P < .001), and ca14 (P < .05), in GF-Katnal2 fish (Fig. S8H–K), suggesting enhanced GABAergic neuronal activity and potential regulatory effects of bacteria on ASD-like zebrafish. These findings proved that the probiotic displayed no different changes on the neuro-related signaling pathways in CR-Katnal2 fish, but the GF-Katnal2 fish showed significant inhibition and rescued by A. muciniphila intervention (Fig. 7B). Furtherly, SCFA content in Katnal2 zebrafish was 104.2 ng/ml, 113.6 ng/ml, 112.2 ng/ml, and 118.2 ng/ml across the four groups, with the GF-Katnal2 + Akk group significantly higher than the CR-Katnal2 fish (Fig. 7C). Expression of the gene ffar2 was used as an indirect readout of SCFA levels: compared with CR-Katnal2, the CR-Katnal2 + Akk group showed a significant increase (Fig. 7C). The GF-Katnal2 group significantly reduced ffar2 expression, whereas GF-Katnal2 + Akk zebrafish exhibited restored ffar2 expression to near-normal CR-Katnal2 levels following A. muciniphila intervention, indicating a partial restorative effect of the probiotic.

Spearman analysis revealed a significant correlation between gut microbiota and host physiological and behavioral traits. Vibrio, Sinorhizobium, Enterococcus, and Aeromonas showed strong positive correlations with DEGs and heartbeats. In contrast, Devoisia and Variovorax exhibited negative correlations with developmental indexes but positive correlations with behavioral ability. Clustering analysis grouped microbiota related to gene expression, movement, and development, suggesting their potential roles in promoting host vitality (Fig. 8A). Redundancy analysis (RDA) by permutation test (P = .073, RDA1 = 50.7%, RDA2 = 27.4%) of key factors revealed that Aeromonas, Enterococcus, and Gemmobacter were positively correlated with heart beats and th, tph1a, ca7, ffar2 expression, while Vibrio and Desulfovibrio were associated with body length, Mycobacterium, and Plesiomonas were closely related to moved distance, moved time, velocity (Fig. 8B). Relationship between microbial features and traits with permutation test indicated marginal significances (P = .068) in groups of CR-AB, CR-AB+Akk, CR-Katnal2, CR-Katnal2 + Akk. Samples within the distinct CR-AB clusters showed a positive correlation with heart rate, whereas those in the CR-Katnal2 clusters were associated with velocity. Key features (e.g. ffar2, th, tph1a, ca7) exhibited sample-specific distributions in CR-AB+Akk and CR-Katnal2 + Akk groups, demonstrating positive regulation of A. muciniphila on host microbial remolding and the subsequent physiological functions (Fig. 8C).

All raw 16S rRNA genes sequencing data supporting this study have been deposited in the NCBI Sequence Read Archive (SRA) and are available under the BioProject accession number PRJNA1433129 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1433129↗).