The probiotic strains are listed in Table 1. The bacteria were stationarily cultivated in MRS medium at 37°C for 24 h. After growing in MRS medium for 24 h, lactobacilli cells were collected by centrifugation (10 min, 4000 × g, 4°C). The pellet was washed by sterile water three times, and resuspended in sterile water at a final concentration of 1.0 × 10⁹ CFU/ml. A. veronii was grown in Luria–Bertani (LB) broth for 18 h at 37°C with 200 rpm shaking.

Adult zebrafish and larvae (Danio rerio) (TU line) were reared in the lab. The adult animals were kept in tanks (length × width × height; 25.5 cm × 18.5 cm × 18.0 cm) in a recirculating aquaculture system under controlled conditions (28 ± 0.5°C, under a 14-h light, 10-h dark photoperiod). The inlet water flow was approximately 1 L/min. The fish were fed twice per day with freshly hatched brine shrimp (8:30 a.m. and 5:30 p.m.). Procedures involving animals were performed in accordance with Chinese legislation associated with animal experimentation and the studies were approved by the Ethics Committee of the Feed Institute, Chinese Academy of Agricultural Sciences (2016-ZZG-ZF-001).

Probiotic strains were grown stationarily in MRS medium at 37°C for 24 h. Bacteria were then pelleted and washed twice in sterile water, and resuspended in water at a final concentration of 1.0 × 10⁷ CFU/ml. At 3 days post fertilization (dpf), about 12 h before hatching, zebrafish eggs were put in contact with the probiotic strains by transferring them to probiotic-containing bottles (60 eggs per bottle). At 4 dpf, the water was exchanged, and fresh probiotic bacteria cells were added. At 7 dpf, fish were infected with virulent A. veronii at a dose of 2 × 10⁷ CFU/ml after water renewal. The mortality was recorded for 5 days.

The protocol to generate and rear GF zebrafish was descripted by Pham et al. (2008) with slight modifications. Freshly fertilized zebrafish eggs were washed by sterilized water three times in a 90 mm sterilized dish, and then the eggs were separated into 50 ml Falcon tubes (100 eggs per tube). Eggs were treated with AB-GZM (gnotobiotic zebrafish medium with antibiotics, which contained 250 ng ml^-1 of amphotericin B, 5 μg ml^-1 of kanamycin, 100 μg ml^-1 of ampicillin, and 10 U⋅ mL^-1 of penicillin and streptomycin) for 4.5 h at room temperature. Then the eggs were washed three times with AB-GZM, and treated with 0.05% of PVPI (polyvinyl pyrrolidone-iodine complex) for 35 s and washed three times with GZM. Next, they were bleached (0.002%) for 15 min. Eggs were washed again three times in GZM and transferred to Petri dishes to be distributed into 300 ml culture bottles with vented caps containing 150 mL of GZM (60 eggs/bottle). GF animals were monitored for sterility every day by spotting 100 μL from each flask on tryptic soy medium agar plates at 28°C under aerobic or anaerobic conditions. Before the gut microbiota transplantation, adult zebrafish were reared in the control or L. casei BL23-added water (1.0 × 10⁶ CFU/ml) for two weeks. Then the fish of the control or L. casei BL23 treatment group were sacrificed and the intestinal contents of every five fish from each group were pooled. Three replicate bottles of GF zebrafish larvae (4 dpf, n = 60) were transplanted with gut microbiota of control or BL23-treated fish at a dose of 10⁵ and 10⁶ CFU/mL. After 3 days of colonization, fish were challenged with A. veronii at a dose of 2.0 × 10⁷ CFU/mL after GZM was renewed. The zebrafish mortality was observed for 120 h after infection.

Total bacteria DNA was extracted from intestinal contents samples by using Power Fecal^TM DNA Isolation kit (MO BIO Laboratories, Carlsbad, CA, United States) according to manufacturer’s instruction. Sequencing was performed at the Novogene Bioinformatics Technology Co., Ltd. Briefly, DNA was amplified by using the 515F/806R primer set (341F: 5′-CCTAYGGGRBGCASCAG-3′ 806R: 5′-XXXXXXGGACTACHVGGG TWTCTAAT-3′), which targets the V-V43 region of the bacterial 16S rDNA, with the reverse primer containing a 6-bp error-correcting barcode unique to each sample. PCR reaction was performed using phusion high-fidelity PCR Mastermix (New England Biolabs LTD., Beijing, China) with the following condition: 94°C for 5 min (1 cycle), 94°C for 20 s/55°C for 20 s/72°C for 30 s (30 cycles), and a last step of 72°C for 10 min. PCR products were purified by using the QIAquick Gel Extraction Kit (QIAGEN, Dusseldorf, Germany). Pyrosequencing was conducted on an Illumina HiSeq 2500 platform according to protocols described by Caporaso et al. (2012). Paired-end reads were merged using FLASH¹ (V1.2.7) (Magoc and Salzberg, 2011). Sequences were then demultiplexed and quality filtered using the default parameters of the Quantitative Insights into Microbial Ecology (QIIME) software package (Caporaso et al., 2010). The operational taxonomic unit (OTU) clustering pipeline UPARSE was used to select OTUs at 97% similarity (Edgar, 2013). The representative sequence sets were aligned and given a taxonomic classification using Ribosomal Database Project (Wang et al., 2007). The similarity among microbial communities was determined using histograms, UniFrac principal coordinates analysis (PCoA), and the unweighted pair-group method with arithmetic mean (UPGMA).

Lactobacillus casei BL23 cells were grown stationarily in MRS medium at 37°C for 24 h. Bacteria were then pelleted and washed twice in sterile water, and resuspended in sterile water at a final concentration of 1.0 × 10⁹ CFU/ml. To obtain dead cells, the bacterial pellets were treated with 4% paraformaldehyde for 2 h, and washed three times in sterile water. After hatching (4 dpf), conventional or GF zebrafish were exposed with live or dead cells of L. casei BL23 at a dose of 1 × 10⁵ CFU/ml, 1 × 10⁶ CFU/ml and 1 × 10⁷ CFU/ml, respectively for 3 days. After three days of probiotics exposure, fish were then infected with virulent A. veronii at a dose of 2 × 10⁷ CFU/ml after GZM was renewed. The mortality was recorded for 5 days. At the infected time of 0, 24, and 48 h, thirty fish (GF and fish treated with live cells or dead cells of L. casei BL23 at the dose of 1 × 10⁶ CFU/ml) from each culture bottle were sacrificed and the whole body was sampled. The samples were immediately frozen in liquid nitrogen and stored at -70°C for cytokine expression analysis.

The method to extract EPSP was previously described by Zhang et al. (2016) in our lab. L. casei BL23 cells were grown stationarily in MRS medium at 37°C for 48 h. Bacteria were then pelleted and washed twice in sterile water, and were resuspended in sterile water. Then bacteria were incubated in water bath incubation (70°C, 24 h). The extracts were precipitated by gradually adding cold ethanol to 75% (v/v), and the supernatant was removed after 24 h, followed by centrifugation at 12000 rpm for 20 min. The precipitated product was washed and dissolved in water obtained from an Alpha-Q reagent grade water purification system (Millipore Co., Milford, MA, United States). The aqueous solution of the extracts were further treated with sevage reagent (trichloromethane – n-butanol, 4:1, vol/vol) at a final concentration of 25% and incubated for 2 h under gentle agitation and then precipitated proteins were removed by centrifugation at 8000 g for 20 min (repeat this step for two times). After centrifugation, the solution containing EPS was dialyzed (molecular weight cut-off: 3000 Da) against 5 l of distilled water for 2 days with water changes three times per day. The extract solution after dialysis was lyophilized.

The purity of the extract (5 mg/ml) was tested by SDS-PAGE electrophoresis and size-exclusion chromatography (SEC) on a column of Superdex75 (10/300 GE) (Pharmacia, Uppsala, Sweden), which fitted to an AKTA FPLC system (Pharmacia) and were eluted with 0.3 M NaCl buffer.

The monosaccharide composition was determined by TLC. Briefly, 20 mg EPSP was hydrolyzed with 2 ml sulfuric acid (1 mol/l) at 100°C for 4 h. The residual sulfuric acid was removed by neutralization with excessive BaCO₃ reaction for 12 h. This solution was adjusted to pH7 and diluted to 20 ml. The hydrolyzate was evaporated under reduced pressure, dissolved in 2 ml ultra-pure water. The resulting hydrolyzate was analyzed by TLC. TLC analysis was conducted according to the previous report (Tanaka et al., 1999). Migration was performed twice on a silica gel TLC plate (20 cm × 20 cm) using n-butanol–methanol–25% ammonia solution–water (5:4:2:1 [vol/vol/vol/vol]). Carbohydrates were visualized by heating the TLC plate after spraying with aniline-diphenylamine reagent (4 ml of aniline, 4 g of diphenylamine, 200 ml of acetone, and 30 ml of 85% phosphoric acid). Monosaccharides of α-Rha, α-Glc, β-GlcNAc, and β-GalNAc were used as standard and the plate was baked at 110°C for 5 min.

The cell line of ZFL was purchased from American Type Culture Collection (ATCC). The ZFL cells were cultured at 28°C in modified limit dilution factor (LDF) culture medium. The complete medium consisted of 50% Leibovitz’s L-15 (L-15), 30 % Dulbecco’s modified Eagle’s (DMEM), and 20% Dulbecco’s Modification of Eagle’s Medium/Ham’s F-12 medium (DMEM/F12) supplemented with 5% fetal bovine serum (FBS), 0.5% trout serum, 10 μg/ml bovine insulin, 50 ng/ml mouse Epidermal Growth Factor (EGF). Mediums were supplemented with 1% penicillin–streptomycin. All basal mediums and FBS were obtained from Corning (NY, United States). The cells were treated with EPSP (10 μg/ml) or equal volume of dd H₂O after the cells covered the plate, and cells were harvested at 24 h after treatment.

Total RNA was isolated from zebrafish larvae and ZFL cells with TRIzol (Invitrogen) extraction. First-strand complementary DNA synthesis was performed using the Superscript First-Strand Synthesis System (Invitrogen). Quantitative real-time PCR reaction were performed using the Power SYBR Green PCR Master Mix (Applied Biosystems) on an ABI 7500 (Applied Biosystems) with reaction volumes of 20 μl. The reaction mixtures were incubated for 5 min at 95°C, followed by 40 cycles of 20 s at 95°C, 20 s at 60°C and 20 s at 72°C, and finally the melt curve was performed from 65to 95°C with a 0.5°C increment for 10 s. Two genes, including rpl13 and rps11 were used as references. The primer sequences are listed in Table 2.

Animal survival rates were analyzed by Kaplan–Meier survival estimate with Bonferroni post hoc test with GraphPad Prism version 5.0 software. Other data were analyzed by one-way analysis of variance (ANOVA) followed by Bonferroni post hoc test with GraphPad Prism version 5.0 software. In addition, unpaired t-test was used to compare data from two groups when appropriate. Wherever applicable, P-values are reported, and a P-value of ≤0.05 is considered significant.

In order to screen the probiotic strains that protect zebrafish larva from damage induced by A. veronii infection, we pre-colonized unhatched (3 dpf) conventional zebrafish larvae with 17 Gram-positive bacteria commonly often used as probiotics in aquaculture and elsewhere in the food industry (Table 1). These pre-colonized larvae were then infected at 7 dpf with A. veronii and their mortality rate was compared to that of the control larvae. The result showed that pre-incubation with Lactobacillus casei BL23 significantly increased the survival rate of larvae upon A. veronii infection (Supplementary Figure S1).

To characterize the protective effect of L. casei BL23 in zebrafish larvae, we first determined whether the protective effect towards zebrafish larvae was dose-dependent. As shown in Figures 1A,B, zebrafish larvae pre-colonized with increasing dosage of L. casei BL23 correlated with increased larvae survival rate after A. veronii infection. No significant difference in survival rate was observed between the larvae pre-colonized with L. casei BL23 at the dose of 1.0 × 10⁵ CFU/ml and control larvae after A. veronii infection (Figures 1A,B). However, the survival rate of larvae pre-colonized with BL23 at 1.0 × 10⁶ CFU/ml or 1.0 × 10⁷ CFU/ml was significantly higher than that of control larvae after A. veronii infection (Figures 1A,B, P < 0.05). These data indicated that the protective effect of L. casei BL23 in zebrafish was dose dependent.

In order to determine whether the protective effect of L. casei BL23 on zebrafish is affected by the bacterium itself or via alterations in the gut microbiota, a GF zebrafish and gut microbe transplant model was established. Here, the intestinal microbiota associated with administration of control treatment and L. casei BL23 treated (L. casei BL23 administration for 2 weeks at a dose of 1.0 × 10⁶ CFU/ml) zebrafish were transferred to freshly hatched GF zebrafish at 4 dpf. These gut microbiota recipient larvae were then infected with A. veronii at 7 dpf and their mortality was compared to that of the A. veronii-infected GF larvae. The results showed survival to be significantly higher in zebrafish colonized with microbiota from either the control or fish treated with L. casei BL23 at doses of 1.0 × 10⁵ CFU/ml and 1.0 × 10⁶ CFU/ml compared with the GF fish (Figures 2A,B, P < 0.001). No significant difference in survival rate was observed between larvae colonized with microbiota from control or BL23-treated zebrafish (Figures 2A,B).

High-throughput sequencing with the 16S rRNA gene was performed to characterize the gut microbiota of zebrafish treated with the control or L. casei BL23. The results showed that phyla Fusobacteria and Proteobacteria and genus of Cetobacterium and Aeromonas were dominant in the intestines of zebrafish (Figures 2C,D). The gut microbial community exhibited no statistical difference between the two groups at the level of phylum to genus (Figures 2C,D and Supplementary Figures S2A–C). These data indicated that administration of L. casei BL23 did not alter the gut microbiota in zebrafish, which is consistent with the results of the microbiota transfer experiment described above.

We then mono-colonized GF zebrafish larvae (4 dpf) with L. casei BL23, then the larvae were infected with A. veronii at 7 dpf. We found that the survival of BL23-treated GF larvae was significantly higher compared with the GF larvae after challenge with A. veronii. The protection mediated by BL23 was found to be dose-dependent (Figures 2E,F, P < 0.001). Among larvae treated with L. casei BL23, the transcription levels of several chemokines, specifically interleukin-1β (IL-1β), tumor necrosis factor α (TNF-α), interleukin-10 (IL-10), and serum amyloid A (Saa), increased 4–80 fold at 24 h after A. veronii infection, but it decreased to near basal levels at 48 h (Figure 3). In contrast, the transcription level of these cytokines in GF fish increased slowly and the increasing trend was maintained for 48 h post challenge (Figure 3). The mRNA levels of TNF-α, IL-1β, IL-10, and Saa were higher in L. casei BL23-treated larvae than in GF larvae at 24 h (Figures 3A–D, P < 0.01) and lower (except IL-10) at 48 h after challenge with A. veronii (Figures 3A,B,D, P < 0.05).

Collectively, these data indicated that the protection of L. casei BL23 against A. veronii infection in zebrafish larvae was mediated by baceterium itself and did not involve the microbiota.

In order to determine whether protective effect of L. casei BL23 in zebrafish is mediated by cellular metabolites or cell structural components, we tested the effect of live and dead (4% paraformaldehyde fixed) cells of L. casei BL23 in both conventional and GF zebrafish. The results showed that, for both conventional and GF zebrafish, live and dead cells of L. casei BL23 both efficiently increased the survival rate of zebrafish larvae after infection with A. veronii (Figures 4B–D, 2E,F, P < 0.05). In addition, a similar difference in cytokine expression was observed between dead BL23-treated larvae and live BL23-treated larvae after A. veronii infection (Figures 4E–H). These data suggest that the immuno-regulation and anti-infectious activity of L. casei BL23 was mediated by certain cell components irrespective of cell viability.

Using the data given above, we speculated that the protective effect of L. casei BL23 in zebrafish might involve the EPS of the bacteria. So, we extracted and tested the biological roles of the EPS from L. casei BL23. GF larvae (4 dpf) were treated with 2, 10, and 20 μg/ml EPS from L. casei BL23, respectively. Then the larvae were infected with A. veronii at 7 dpf, and the survival rates of these fish were compared to the A. veronii-infected GF fish. The survival rate of fish treated with the L. casei BL23 EPS at 10 and 20 μg/ml was significantly higher than those of GF fish after A. veronii infection (Figures 5A,B, P < 0.01).

The EPS extract was purified by SEC on a column of Superdex75 (10/300 GE). Interestingly, as shown in Figure 6A, the EPS extract showed a single, symmetrical protein peak, indicating that the EPS extract contains a homogeneous protein. It was also confirmed by the SDS-PAGE analysis, which showed a single band with 40–45 KD in size (Figure 6B). In addition, qualitative carbohydrate analysis of the fractions (purified by SEC) with the phenol–sulfuric acid method showed that all the positive reactions were within the peak of the protein fractions. Furthermore, the monosaccharide composition of the EPS extract was determined by TLC after acid hydrolysis. TLC indicated that the EPS of the cell extract from L. casei BL23 was composed of α-Rha, α-Glc, β-GlcNAc, and β-GalNAc (Figure 6C). Together, these findings indicated that the EPS extract is an exopolysaccharide-protein complex (EPSP).

In order to characterize EPSP-mediated immuno-regulation in zebrafish, we tested the immune response of ZFL cells after EPSP treatment. The transcription level of toll-like receptors (TLRs) and cytokines were evaluated in ZFL cells at 24 h after EPSP treatment. The results showed that EPSP induced more expressionl of TLR1 and TLR2 compared with the control (Figures 7A,B, P < 0.05). In addition, the transcription level of IL-10 and TNF-α was significantly higher in ZFL cells treated with EPSP than in the control (Figures 7C,D, P < 0.05). Gene expression of IL-1β was less pronounced than in the control after EPSP treatment in ZFL cells (Figure 7E, P < 0.05). No significant difference between the EPSP-treated and control ZFL cells was observed in the transcription levels of TLR3, TLR4a, TLR5a, TLR5b, MyD88, NF-κB, or IL-6 (Figure 7F and Supplementary Figures S3A–G). These data suggested that the immuno-regulation of EPSP on zebrafish were might involve the TLR1/TLR2 signal pathway.