V. alginolyticus ATCC33787 was purchased from the Guangdong Province Microbial Culture Collection (GDMCC). V. alginolyticus is grown in 3% NaCl 0.5% yeast overnight and 1:100 using fresh 3% NaCl 0.5% yeast medium and grown at 30°C. For growth curve, OD600 of the bacterial cultures was measured in medium with 0, 0.78, 3.125, 12.5, 50 or 200 mM MgCl₂, which were designed according to the event that 50 mM is approximately the MgCl₂ concentration of coastal waters and then 50 time or divided by 4 for other concentrations. Notably, different from V12G01 used in a previous report (36), which caused infectious symptoms to significantly appear after 40 h, ATCC33787 led to the similar infectious symptoms after 24 h.

V. alginolyticus was inoculated into 5 ml of 3% NaCl and 0.5% yeast medium and cultured at 30°C overnight. Then 10 μl of each sample was spotted onto the center of 1.5% or 0.3% agar LB plates with 3% NaCl and 0, 0.78, 3.125, 12.5, 50 or 200 mM MgCl₂ for swarming and swimming, respectively. These plated samples were incubated in a constant temperature incubator at 30°C for 8 h for the diameter of the halo.

Zebrafish (~0.2g body weight), Danio rerio, were obtained from a zebrafish breeding corporation (Guangzhou, China). Zebrafish were reared in 25 L water tanks and each tank was equipped with closed recirculating aquaculture systems. These fish were cultured for two weeks before experimental manipulation and were fed twice daily.

For the metabolic profile of zebrafish injected by vaccine, zebrafish (n = 180) were randomly divided into two groups, control group (n = 90) and bacterial vaccination group (n = 90). The control group was injected with 5 μl of 3% saline per fish and the vaccination group was intramuscularly injected with 5 μl V. alginolyticus cultured with 200 mM MgCl₂ (2 × 10⁵ CFU) per fish. Spleens were collected at 48 h post-injection. Nine spleens were mixed for a sample. A total of ten samples were obtained for GC–MS analysis.

For the metabolic profile of survival and dying zebrafish from bacterial infection, zebrafish (n = 270) were randomly divided into two groups, control group (n = 90) and bacterial challenge group (n = 180). The control and bacterial challenge groups were intramuscularly injected with 5 μl of 3% saline and 5 μl V. alginolyticus (8 × 10⁵ CFU, half lethal dose) per fish, respectively. Spleens were collected at 24 h post-injection. Nine spleens were mixed for a sample. A total of ten samples were obtained for GC–MS analysis.

To prepare the fish sample for the GC–MS analysis, samples were quenched with 1 ml of cold methanol (HPLC, Sigma Aldrich) and sonicated for 5 min at a 200 W power setting. After centrifugation at 12,000g for 10 min, 10 μl of 0.1 mg/ml ribitol (Sigma Aldrich) as an analytical internal standard was added. The supernatant was concentrated for 4 h in a rotary vacuum centrifuge device, LABCONCO. The dried polar extracts were used for the GC–MS analysis. Ten biological samples with two technical repeats were separately used for the test and control groups.

GC–MS analysis was carried out with a variation on the two stage techniques as described previously (37, 38). In brief, samples were derivatized through two steps. First, 80 μl of 20 mg/ml methoxamine hydrochloride in pyridine (Sigma) was added to the extracts and incubated for 3 h at 37°C. Then, 80 μl N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA, Sigma-Aldrich) was put and incubated for 1.5 h at 37°C. The samples were centrifuged at 12,000g for 10 min at 4°C. A sample analysis was carried out by Agilent 7890A GC equipped with an Agilent 5975C VL MSD detector (Agilent Technologies, Santa Clara, CA, USA). The injector temperature was kept at 270°C, and 0.1 μl aliquot was injected into a column. Temperature program of the GC oven was held at 85°C for 5 min, followed by an increase to 270°C at a rate of 15°C min and then held for 5 min. Helium was used as carrier gas and its flow rate was 1 ml/min. The MS was operated in a range of 50–600 m/z. For each sample, two technical replicates were prepared to confirm the reproducibility of the reported procedures.

To investigate the effect of the TCA cycle on host survival, a total of 80 zebrafish were randomly divided into saline group (n = 20) and sodium malonate group (n = 60). For sodium malonate group, the fish were further randomly divided into three subgroups (n = 20). Each subgroup was reared in an individual tank. Control group was injected with 5 μl saline (0.85% sodium chloride) per fish, while the three subgroups were separately injected with 12.5, 25, and 50 mM sodium malonate, once per day for three days via intraperitoneal injection. These fish were intramuscularly challenged with 5 μl V. alginolyticus (8 × 10⁵ CFU, half lethal dose) per fish.

The activity of pyruvate dehydrogenase (PDH), a-ketoglutarate dehydrogenase (a-KGDH), succinate dehydrogenase (SDH) and malate dehydrogenase (MDH) was measured as previously reported with a modification (39). Briefly, visceral organs from three zebrafish were pooled and homogenized in ice-cold PBS (pH 7.4) which were added at a ratio of 1:15 (w/v). The homogenates were centrifuged and protein concentration in supernatant was measured by a BCA Protein Assay Kit (Beyotime, China). The reaction buffer for PDH and a-KGDH included 0.5 mM MTT, 2.5 mM MgCl₂, 6.5 mM PMS, 0.2 mM TPP, 50 mM PBS, and 2 mM sodium pyruvate (for PDH) or 2 mM sodium α-Ketoglutaric acid (for α-KGDH). The reaction systems of SDH and MDH included 0.5 mM MTT, 13 mM PMS, 50 mM PBS, 5 mM succinate (for SDH) or 50 mM malate (for MDH). All the reactions were performed in a final volume of 200 μl in a 96-well plate. Subsequently, the plate was incubated at 30°C for 10 min for PDH, a-KGDH, SDH, and MDH. The optical absorbance was performed in a microplate reader (Bio-Tek Synergy2, USA) at 566 nm.

For expression of genes encoding the TCA cycle, zebrafish (n = 252) were divided into five groups, control group (n = 24), vaccination group (n = 24), L group (n = 24), M group (n = 60), and H group (n = 120). Control and vaccination groups were intramuscularly injected with 5 μl saline solution and 5 μl V. alginolyticus cultured with 200 mM MgCl₂ (2 × 10⁵ CFU), respectively, each fish. L group, M group, and H group were intramuscularly injected with 5 μl V. alginolyticus (2 × 10⁵ CFU, 8 × 10⁵ CFU, 1.5 × 10⁶ CFU, respectively) each fish. These zebrafish were collected at 24 h post-injection. For expression of genes encoding innate immune immunity, zebrafish were intramuscularly injected with 5 μl (50 or 100 μg) fumaric acid, once a day for three days, and collected from vaccination and bacterial challenge as described earlier. Expression of these genes was analyzed by the qRT-PCR as previously described (39). The total RNA was isolated from spleens pooled from six D. rerio with Trizol (Invitrogen, USA). qRT-PCR was performed in 384-well plates with a total volume of 10 μl containing 5 μl 2 × SYBR Premix Ex TaqTM, 4.6 μl H₂O, 0.1 μl cDNA template and 0.2 μl each of forward and reverse primers (10 mM). The reaction mixtures were run on a LightCycler 480 system (Roche, Germany). Data were shown as the relative mRNA expression compared with the β-actin gene by 2^−ΔΔCT method. All qRT-PCR reactions were performed for four biological replicates. Gene-specific primers used for qRT-PCR are shown in Table 1 .

To investigate the effect of magnesium ion on physiology of V. alginolyticus, growth curve of V. alginolyticus ATCC33787 was determined in a gradient concentration of MgCl₂. Different growth rate of the bacterium was observed in different MgCl₂ concentrations, especially during the first 1–2 h. The growth rate of ATCC33787 from the fastest to the slowest were in 50 mM ≥ 12.5 mM > 3.125 mM > 0.78 mM > 0 mM > 200 mM MgCl₂ ( Figure 1A ). Furthermore, swarming and swimming experiments were also carried out in the gradient concentration of MgCl₂. The strongest swarming was measured in plates with 0.78 mM MgCl₂ and then reduced with the increasing MgCl₂ ( Figure 1B ). Interestingly, the swarming ability was lower in plate with 200 mM MgCl₂ than without MgCl₂ ( Figure 1B ). However, MgCl₂ concentration did not affect swimming ability in plates with 0.78–50 mM MgCl₂ and without MgCl₂, where the bacterium diffused to all the plates but not evenly distributed in 50 mM MgCl₂. When the concentration of MgCl₂ reached 200 mM, the distribution was depressed ( Figure 1B ). Importantly, virulence of ATCC33787 to zebrafish was reduced with the increasing MgCl₂ concentrations, where similar survival was detected in zebrafish infected by bacteria cultured in 200 mM MgCl₂ and saline control ( Figure 1C ). When zebrafish were infected with different numbers of 200 mM MgCl₂-cultured bacteria, percent survival of zebrafish was elevated with the decreasing numbers of bacteria. Among them, all animals were survived at the infection dose of 2 × 10⁵ CFU bacteria ( Figure 1D ). Furthermore, zebrafish were immunized with 2 × 10⁵ CFU of 200 mM MgCl₂-cultured bacteria or saline. Then, zebrafish were challenged with a lethal dose of 5 × 10⁶ CFU V. alginolyticus ATCC33787. When all animals died in the control group, 76% animals survived in the experiment group ( Figure 1E ). These results indicate that 200 mM MgCl₂-cultured ATCC33787 can be used as a live-attenuated vaccine candidate against infection caused by V. alginolyticus.

Hosts against bacterial infection have infective and anti-infective metabolomes, which decide the consequences of infection (36, 40). This motivated us to explore the metabolic mechanisms by which the vaccine candidate protects zebrafish against the bacterial infection. To test this idea, GC–MS based metabolomics was adopted to investigate the metabolic profiles of control group (injected with saline) and vaccination group (injected with 2 × 10⁵ CFU of the V. alginolyticus vaccine). Nine spleens were pooled as one sample. A total of 240 aligned peaks were identified in each sample. After the removal of internal standard, ribitol, any known artificial peaks, and merge of the same compounds, 80 metabolites with reliable signals were identified in each sample. Ten samples with two technical repeats for each sample were included in each group, yielding 40 data sets. The correlation coefficient between technical replicates varied between 0.996 and 0.999, demonstrating the reproducibility of the data ( Supplementary Figure 1A ). According to the Kyoto Encyclopedia of Genes and Genomes (KEGG), 38, 27, 20, 12, and 3% of the metabolites were categorized to carbohydrate, amino acid, fatty acid, nucleotide, and others, respectively ( Supplementary Figure 1B ). The metabolomic profiles of these two groups were displayed as heat map ( Supplementary Figure 1C ). Kruskal–Wallis test was used to compare the two groups, where 61 differential abundances of metabolites were identified as shown in Figure 2A . A Z-score plot ranged from −4.9 to 18.1 in the vaccination group compared to control group and showed 15 increased metabolites and 46 decreased metabolites ( Figure 2B ). Among these differential abundances of metabolites, 37, 34, 17, and 12% of the metabolites were carbohydrate, amino acid, lipid, and nucleotide ( Supplementary Figure 2 ). These differential abundances of metabolites were analyzed and outlined in KEGG (http://www.genome.jp/kegg) and MetPA (http://metpa.metabolomics.ca), respectively, for pathway enrichment. A total of eight metabolic pathways were enriched, where the top three impactful pathways were TCA cycle, pyruvate metabolism, alanine, aspartate and glutamate metabolism ( Figure 2C ). Interestingly, among the eight enriched pathways, all differential abundances of metabolites were elevated in the TCA cycle and pyruvate metabolism but reduced in valine, leucine, and isoleucine biosynthesis ( Figure 2D ). Therefore, the TCA cycle can be further explored for vaccine-conferred protection.

To understand differential metabolic profiles between survival and dead zebrafish infected with V. alginolyticus, GC–MS was used to characterize the metabolomes of survival and dying zebrafish infected with LD50 of ATCC33787 and saline group. GC–MS was performed as described above. The identified 72 metabolites (p <0.05) in survival and dying groups were shown as heatmap, where survival and control groups were clustered ( Figure 3A ). The differential abundances of metabolites were identified between the control and survival groups or control and dying groups by Kruskal–Wallis test. Approximately 59 metabolites and 64 metabolites were identified (p <0.05) in survival metabolome and dying metabolome, respectively. Z scores value of these differential metabolites showed that fumaric acid was the most upregulated metabolite in the survival group, and malic acid was the most downregulated substance in the death group ( Figures 3B,C ). Pathway analysis of 59 differential metabolites in survival group and 64 differential metabolites in dying group were performed in MetPA, fourteen and thirteen pathways were enriched, respectively, as shown in Figures 3D, E . Interestingly, all metabolites of the TCA cycle and glycine, serine and threonine metabolism were elevated ( Figure 3F ). These results indicate that the elevated TCA cycle and glycine, serine and threonine metabolism are required for the survival.

It is interesting to compare the metabolomes induced by the vaccine and mediated by the infection. To do this, a Venn diagram was used to exhibit overlapped and unique metabolites among 61, 59, and 64 differential abundances of metabolites of the vaccination group, survival group, and dying group, respectively. Forty metabolites were shared among these three groups; 6, 10, and 10 were overlapped between the survival and vaccination groups, survival and dying groups, vaccination and dying groups, respectively; 3, 5, and 4 existed only in the survival, vaccination and dying groups, respectively ( Figure 4A ). Among these 14, 26, and 31 upregulated metabolites of the vaccination, survival and dying groups, respectively, 2 metabolites were shared by the three groups; 7, 10, and 1 metabolites were shared by the survival and vaccination groups, survival and dying groups, vaccination and dying groups, respectively; 7, 4, and 18 metabolites were only present in the survival, vaccination and dying groups, respectively ( Figure 4B ). Among the 47, 33, and 33 downregulated metabolites of the vaccination, survival and dying groups, respectively, 12 metabolites were shared by the three groups; 10, 9, and 5 were shared by the survival and vaccination group, survival and dying group, vaccination and dying group respectively; 2, 20, and 7 were only present in the survival, vaccination and dying groups, respectively ( Figure 4C ). When all of the identified metabolites were analyzed together among the three groups, the vaccination and survival groups were clustered ( Figure 4D ), indicating the survival and vaccination groups have similar metabolic profiles.

Moreover, orthogonal partial least squares discriminant analysis (OPLS-DA) was adopted to identify shared biomarkers between vaccination and survival groups. To do this, S-plot was used for identification of discriminatory variables. Cut-off values were set as greater or equal to 0.5 and 0.05 absolute value of covariance t and correlation p (corr), respectively. As compared to the control group, seven metabolites were increased and eight metabolites were decreased in the vaccination group ( Figure 4E ). Whereas five metabolites were increased and 10 metabolites were decreased in the survival group ( Figure 4F ), while eight metabolites were increased and 10 metabolites were decreased in the dying group ( Figure 4G ). Among the shared metabolites, the abundance of maltose, glucose, taurine was upregulated and the abundance of glycolic acid, myo-inositol, stearic acid, palmitic acid was downregulated in both of the survival and vaccination groups. Malic acid, an intermediate metabolite of the TCA cycle, was increased in the survival group and decreased in the dying group ( Figures 4E – G ). PCA analysis among the survival group, dying group and vaccination group further confirms this conclusion. Compared to the dying group, pentadecanoic acid, malic acid, cystathionine, ethanimidoate, glucose, mandelic acid, maltose, galactose, benzoic acid and glutamic acid were increased; myo-inositol, threonine, threonic acid, thymine, threose, palmitelaidic acid, leucine, serine, uridine, heneicosanoic acid, 1-monolinoleoyllglycerol, 4-hydroxybutyric acid, valine, creatinine, heptadecanoic acid, maleic acid, cytosine, stearic acid, palmitic acid, glycolic acid and beta-lactic acid were decreased in the survival and vaccination groups ( Figure 4H ). These results indicate that the metabolic flux of glycolysis to the TCA cycle instead of fatty acid biosynthesis is shared by the vaccination and survival groups.

The above results suggest that the TCA cycle plays a crucial role in the protection against bacterial infection. To confirm this, the activity of pyruvate dehydrogenase (PDH), α-Ketoglutarate dehydrogenase (KGDH), succinate dehydrogenase (SDH) and malate dehydrogenase (MDH) in the TCA cycle and pyruvate metabolism were measured. The activity of all enzymes was elevated in the vaccination group, survival group and reduced in the dying group ( Figure 5A ). Then, iPath was used to compare metabolic pathways among the vaccination, survival and dying groups. The resulting global overview map provided a better insight into the effects of vaccination and infection consequence on the metabolism of the fish, where yellow and blue lines represented increased and decreased pathways in the reprogramming group, respectively. Elevation of the TCA cycle in the vaccination and survival groups and fluctuation of the TCA cycle in dying group form the most characteristic feature ( Figures 5B – D ). Furthermore, qRT-PCR was used to detect expression of genes encoding the TCA cycle of zebrafish challenged by high (1.5 × 10⁶ CFU, H), middle (8 × 10⁵ CFU, M) and low (2 × 10⁵ CFU, L) doses of ATCC33787 and the 200 mM MgCl₂-prepared live-attenuated vaccine (vaccination) and saline solution was used control. They caused 15% survival (survival-H) and 85% dying (dying-H), 60% survival (survival-M) and 40% dying (dying-M), 100% survival (survival-L), and 100% survival (vaccination), respectively. On the whole, higher expression of genes was detected in the three survival groups and vaccination group than the control, while lower expression of genes was measured in the two dying groups than the control. The high expression was ranked as survival-H > survival-M and vaccination > survival-L and low expression was listed as dying-H > dying-M ( Figure 5E ). Finally, sodium malonate, an inhibitor of SDH, decreased the survival of zebrafish infected with ATCC33787 in a dose-dependent manner ( Figure 5F ). These results indicate that elevation of the TCA cycle is required for the survival of zebrafish infected with ATCC33787. Consistently, the elevated TCA cycle is a possible reason why the live-attenuated vaccine provides an effective ability against the infection by the bacterium.

Innate immune response contributes to host immune protection against bacterial infection (41, 42). Therefore, it is required to investigate whether the live-attenuated vaccine activates an innate immune response. The expression of ten innate immune genes il1β, il4, il8, il21, tnf-α, c3b, tlr1, tlr3, nf-κb, and lysozyme were measured. Compared to the saline control, the vaccination group exhibited elevated expression of il1β, il8, and lysozyme but the rest of the genes remain unaffected ( Figure 6A ). Meanwhile, expression of these genes was compared between the survival group and the dying group. The expression of il1β, il8, il21, tnf-α, tlr1, nf-κb, and lysozyme was elevated in the survival group, while the expression of il1β, il4, tlr1, tlr3, and nf-κb was reduced in the dying group ( Figure 6B ). Importantly, the elevated il1β, il8, and lysozyme were shared between the vaccination and survival groups, where expression of il1β was reduced in the dying group.

Recently, we have showed that malic acid of the TCA cycle promotes expression of innate immune genes and thereby improves the survival of zebrafish infected with V. alginolyticus (40). We speculate that the elevated TCA cycle is related to the elevated expression of innate immune genes. To test this, fumaric acid of the TCA cycle was used to test whether it can promote expression of the three shared genes il1β, il8, and lysozyme. Indeed, exogenous fumaric acid increased expression of il1β, il8, and lysozyme in a dose-dependent manner. Besides these, the metabolite also promoted the expression of il21 and nf-κb ( Figure 6C ). These results indicate that the live-attenuated vaccine improves the immune level of zebrafish by promoting the TCA cycle, and thus elevates the survival rate of zebrafish infected with V. alginolyticus.