The E. faecium strain, LAC7.2, which was selected as a probiotic, was isolated from the gastrointestinal tract of healthy Nile tilapias from the hatchery of Londrina/Paraná, Brazil in 2017. In vitro tests were performed to suggest the probiotic potential. Nile tilapia feces were diluted (scale 10) in 0.85% saline, plated on Man, Rogosa, and Sharpe (MRS) Lactobacillus Kasvi^® agar, and incubated for 48 h at 28°C. Thereafter, colonies were selected and characterized. The selected bacteria strains were then seeded on MRS agar and incubated for 48 h. A solution containing Mueller Hinton agar (Kasvi, São José dos Pinhais, Brazil) at 45°C with pathogenic bacteria (Escherichia coli, Staphylococcus spp., and Streptococcus spp.) was prepared. This solution was placed on MRS plates containing probiotic bacteria. The plates were incubated for 24 h at 28°C, and the inhibition halos were measured for bacteria with larger halos. To evaluate the inhibition induced by Francisella noatunensis F1, cystine heart agar enriched with 1% of bovine hemoglobin (Kasvi) was employed, where the pathogenic bacteria were seeded with swabs over the plate surface. Thereafter, small holes were made in the agar and the E. faecium LAC 7.2 filtered supernatant (0.22 μm) was deposited. Readings were taken after 48 h. This test was performed in duplicate. The strain that shown the higher inhibition halo against pathogenic bacteria was considered as a potential probiotic for in vivo trial.

Genome sequencing of E. faecium was performed using the MiSeq platform (Illumina^®, United States). Reads were uploaded in FASTQ format to the CLC Genomics Workbench 12 (Qiagen, United States) software for the trimming and assembly steps. Genome annotation was performed in Rapid Annotation using Subsystem Technology (RAST; version 2.0) (Aziz et al., 2008). The genome project was deposited in the GenBank database under the accession number CP045012.1. A summary of the project information is provided in Table 1.

Phylogenetic analysis was performed using Gegenees V2.2.1 (Ågren et al., 2012) and SplitsTree4 v4.15.1, with high accuracy, to generate a heatmap and phylogenetic tree, respectively. Secondary metabolite clusters in the sequenced genome were predicted using antiSMASH 5.0 (Weber et al., 2015) and RAST. The ResFinder 3.2 program was also used to detect resistance genes.

A total of 405 Nile tilapia (O. niloticus) were obtained from a commercial hatchery in the state of Paraná, with an initial weight of 11.93 ± 0.59 g. The animals were stored in 150 L tanks containing heated water with continuous renewal (80% of daily volume) for an acclimatization period of 7 days; the temperature was maintained at approximately 25°C and fish were fed three times per day until apparent satiety. Water parameters (pH, 6.8–7.2; total ammonia, <0.4 mg/L, dissolved oxygen, 5.4 mg/L; and absence of chlorine) were measured daily and maintained throughout the experimental period. The microbiological diagnosis was performed before the experiment, where 20 fish were randomly sampled and killed using a high dose of benzocaine (200 mg/mL). All animal procedures were approved by the Ethics Committee on Animal Use of the State University of Londrina (CEUA/UEL-7327.2017.39). Fragments of the brain, liver, cranial kidney, and spleen were stripped in Mueller Hinton agar enriched with 5% defibrinated sheep blood and in cystine heart agar enriched with 1% of bovine hemoglobin. The plates were incubated at 28°C for 5 days to confirm the health status of the fish (no bacterial growth in the plates).

Fish were divided into three treatment groups: G1 (probiotic rationing), G2 (probiotic in water), and G3 (probiotic in feed and water); and three control groups: PCA (no probiotic and challenge with Francisella orientalis), PCB (no probiotic and challenge with Streptococcus agalactiae), and NC (no probiotic and no challenge) (n = 45), as shown in Table 2. The fish were fed a commercial feed (Presence^® Nutripiscis TR 36% CP) according to the treatments and challenged on day 38. After infection, the animals were monitored for 30 days to determine clinical signs and mortality.

For the G1 group, approximately three bacterial colonies of E. faecium were added to 600 mL of MRS Lactobacillus Kasvibroth^® and incubated under agitation for 48 h at 28 °C. Thereafter, 100 mL was sprayed on 1 kg of feed (2.3 × 10⁸ CFU/g of feed), with 5 mL of universal vehicle (Vansil^®) and dried at 28°C for 8–12 h. Fish were fed four times per day until satiety. For group G2 (aquarium volume around 27 L), 100 mL of MRS broth with cultured E. faecium was added to aquarium water and the water renovation was interrupted for 2 h. Thereafter, the water volume was restored. This procedure was performed every 10 days. Group G3 received both treatments.

On the day before challenge with F. orientalis, the water temperature was gradually decreased and maintained at 21°C (±1°C) to promote immersion infection for 3 h (7.1 × 10⁵ CFU/mL of water in the tank); this is because outbreaks of franciselose in Brazil occur in colder waters (Ortega et al., 2016). For infection with S. agalactiae, the water temperature was maintained at 28°C (±1°C) (Chen et al., 2012) and administered intraperitoneally at 0.1 mL/fish (8.8 × 10⁵ CFU/mL) schematic figure about isolation and treatment of fish with a probiotic bacteria. Previous data from our team showed that the immersion route is more suitable for francisellosis infection, and it is possible to establish a Lethal Dose concentration of (LD₅₀). For S. agalactiae, only the intraperitoneally route reproduces the disease and it was possible to calculate the LD₅₀ (unpublished data). In Supplementary Figure 1 represents a schematic of the isolation and administration of E. faecium and Supplementary Figure 2 the route of application for the challenge of pathogenic bacteria in fish.

Fish of all replicates were counted and weighed individually on the first and last day of trial. The weight gain, weight gain (%), medium final body weight, and specific growth rate (SGR) were determined (Ridha, 2006).

Blood samples were collected at 38 days post-treatment with probiotics (16 samples per group). The fish were anesthetized with benzocaine (100 mg/L), and blood was collected by puncturing the caudal vessel in 3 mL syringes (21 G) containing 10% anticoagulant (ethylenediaminetetraacetic acid). Blood samples were used to measure hematocrit levels (Hct; %) using the microhematocrit method (Paiva et al., 2013), and red blood cells (RBCs; 10⁶/μL) were counted in a Neubauer chamber following dilution at 1:200 in Dacie solution. White blood cells (WBCs; 10³/μL) and total thrombocyte counts were calculated using an indirect method (Ishikawa et al., 2008). For differential counting of leukocytes, smears were stained with May–Grünwald/Giemsa/Wright stain. The hemoglobin concentration (Hgb; g/dL) was analyzed using the cyanmethemoglobin method (Collier, 1944) using commercial kits (Labtest, Lagoa Santa, MG, Brazil) to determine the hematimetric indices of mean corpuscular volume (MCV; fL) and mean corpuscular hemoglobin concentration (MCHC; g/dL).

Innate immune analysis was performed in all groups (five fish per replicate) at 38 days after treatment. Blood samples were collected without anticoagulant, allowed to coagulate, and centrifuged at 1400 × g for 10 min at room temperature to obtain serum, which was stored at -20°C.

Serum lysozyme activity was determined using a methodology adapted from that described by Demers and Bayne (1997). Briefly, the initial and final absorbances were measured by spectrophotometry to determine the serum lysozyme activity by the lysis of the Gram-positive bacterium, Micrococcus lysodeikticus (Sigma-Aldrich Chemical Co.). The reduction in absorbance of the samples was converted into an estimate of lysozyme concentration (μg mL).

Alternative complement pathway activity (ACH50) was determined using rabbit red blood cells (RaRBCs) as target cells for hemolysis, following a previously described method (Sunyer and Tort, 1995). Briefly, serially diluted sera were mixed with rabbit erythrocyte suspension and incubated at 25°C for 1 h with occasional shaking. The extent of hemolysis was estimated by measuring the optical density of the supernatant at 414 nm (OD₄₁₄). Serum dilutions resulting in greater than 90% or less than 15% lysis were excluded from the calculation, and the serum dilution that resulted in 50% lysis of RaRBC was represented as ACH5O units/μL.

After 38 days of treatment, six fish from each group were used for the bacterial microbiome analysis, and each DNA sample was isolated from the stools of two fish and pooled. The animals in each experimental group were killed with benzocaine (200 mg/L). The stool of the entire intestinal tract was removed aseptically and maintained in sterile vials with refrigeration. The samples were immediately stored in a freezer at -80°C until processing. For total DNA extraction, a commercial QIAamp DNA Stool Mini Kit (QIAGEN, Hilden, Germany) was used according to the manufacturer’s instructions. Thereafter, the V4 region of the 16S ribosomal subunit gene was amplified with primers containing overlapping regions with Illumina platform primers (Klindworth et al., 2013). After verification of the amplicon quality, the samples were sent to the Neoprospecta company for sequencing using the Illumina MiSeq platform with the 250-cycle V2 kit. The next steps were performed according to Suphoronski et al. (2019) using MOTHUR v.1.36.1 software.

Data were subjected to normality and homogeneity tests and subsequently to analysis of variance (ANOVA), followed by the Tukey test for comparison between arithmetic means, with a significance level of 5%. For quantitative variables that did not present a normal distribution, the non-parametric Kruskal–Wallis test was used, followed by the Dunn test with a significance level of p < 5%. Cumulative mortality was analyzed using the Fisher exact test with a significance level of 5% using OpenEpi v. 3.01.¹ In metagenomics, to verify the abundance significance of taxa between groups, statistical analysis of metagenomic profiles (STAMP) was performed using parent level 1 and profile level 6 to analyze the significance between two groups using the two-sided Welch’s t-test (Parks et al., 2014).

The inhibition halos for the selected strains were 16 mm (E. coli), 15 mm (Staphylococcus spp.), and 10 mm (Streptococcus spp.) (data not shown). The inhibition zone of F. orientalis F1 was 18 mm. The E. faecium antagonism against Staphylococcus spp. and the inhibition of supernatant (filtered or not) of probiotic strain are demonstrated in Supplementary Figure 3. The complete genome of the E. faecium strain consisted of a single circular chromosome that is 2.625.745 bp in length and two plasmids of 206.375 and 80.816 bp, totaling 2.912.936 bp, 37.96% G + C content, 18 rRNA operons, 68 tRNA genes, and 131 pseudogenes (Table 1). In the phylogenetic tree, the ancestors of the bacterium were found to be Enterococcus casseliflavus (Figure 1). Based on the heat map, E. faecium was highly similar to other strains of the same species, but was not 100% similar to other strains. In the anti-SMACH analysis, the secondary metabolite, Enterocin A (Access GenBank: QIS84411.1 and locus_tag F6447_10905), was found to have 100% similarity in region 10 while polysaccharide had 82% similarity in region 3 (Figure 2). In ResFinder, two resistance genes were found in one of the plasmids: aminoglycosides and macrolides.

The mean probiotic concentration was 5.53 × 10⁶ CFU/g. The final body weight and weight gain of the medium are shown in Table 3. No significant differences (p > 0.05) were observed in weight gain (g) (Table 3); however, the final body weight (g), SGR, and weight gain (%) were significantly higher in the G1 group than in the other groups, including the NC group.

There were no significant differences (p > 0.05) between the treated and control groups in the hematocrit, hemoglobin, erythrocytes, MCV, and MCHC parameters. Thrombocyte counts were significantly larger in G1 than in G2 and similar to G3. Further, there was a difference in the MCH between NC and G2. The differential leukocyte cell counts did not differ significantly between treatments for total leukocytes, neutrophils, or monocytes. No eosinophils or basophils were observed in any of the treatments tested (Table 4). The mean serum lysozyme concentrations and complement activity did not differ significantly between the treatment groups and the NC group (p > 0.05) (Table 4).

A total of 9.038.053 sequences were obtained for all groups, and 309 operational taxonomy units (OTUs) were identified. A rarefaction curve showed that sequencing was sufficient to identify most of the bacterial species present in the fish gut, suggesting that the read count of the trial was representative of the intestinal bacteriome in all groups (Figure 3). Table 5 displays the total number of sequences obtained for each group. Mothur software was used to calculate the Shannon index, which compares the diversity of species in each group. However, no significant differences were observed between groups. The G1 and NC groups presented the lowest level of species diversity, whereas the G2 and G3 groups had a higher level of diversity than the G1 and NC groups. Such finding suggests that groups receiving probiotic in water had greater diversity. The abundance of bacterial species calculated using the Mothur software is shown in Figure 4. The abundance plot displays the most abundant bacteria in each sample: Cetobacterium, Vibrionaceae_unclassified, Plesiomonas, Gammaproteobacteria_unclassified, Streptophyta_unclassified, Bacteroidales_unclassified, and Enterobacteriaceae_unclassified. Cetobacterium was most abundant in all groups; however, the G1 and NC groups had a high percentage of Cetobacterium compared with the G2 and G3 groups. Vibrionaceae_unclassified was more abundant in G2 and G3 than in G1 and NC.

The cumulative mortality of fish infected with F. orientalis F1 was: PC, 65.12%; G1, 68.18%; G2, 62.15%; and G3, 54.35%. No statistical difference was found between the groups (Figure 5). However, the cumulative mortality of fish infected with S. agalactiae S13 showed a statistical difference between the PC and other groups that received probiotics in feed and/or water. Mortality was: PC = 88.29%; G1 = 75.56%; and G2 and G3 = 73.33% for both (Figure 6).