Animal studies involved in this manuscript adhere to the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and maintained according to the standard protocols. All the experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (Approval No. SYSU-IACUC-2020-B126716).

Aeromonas hydrophila LP-2 was a generous gift by Prof. X. M. Lin from Fujian Agriculture and Forestry University. The bacteria were confirmed by 16S rDNA sequencing. The bacteria were stored at -80°C and cultured in Luria-Bertani medium at 30°C. Crucian carp with average weight 3 g ± 0.5 g were obtained from a local crucian carp-breeding corporation (Panyu District, Guangzhou, China) and were free of Aeromonas infection through microbiological detection. The fish were maintained in a 170-l water tank equipped with a closed recirculating aquaculture system and fed once a day with commercial fish food. The fish culture room was equipped with normal light at a rhythm of 12-h/12-h light and darkness.

For bacterial infection, a single colony of A. hydrophila was picked up from the agar plate and grown in LB medium overnight at 30°C shaken at 200 rpm. The overnight culture was then reinoculated to fresh LB medium at a ratio of 1:100 and grown with shaking at 200 rpm until OD600 reached 1.0. Cells were pelleted by centrifugation, washed with 0.85% sterilized saline solution for three times, and resuspended in the same solution.

For bacterial infection, Carassius carassius were individually challenged with 5 μl of 1 × 10⁶ CFU bacterial suspension or saline only via intraperitoneal injection (n = 20 for each treatment). Fish mortality was examined twice daily for 10 days to monitor accumulative death. The dying crucian carp were selected when they had the following symptoms: darkening body, slow and unbalanced swimming, and swimming close to the water surface; the crucian carp that survived after 3 days, were swimming smoothly, and had a normal body color were defined as survived fish.

The sample preparation was carried out as previously described (32, 47). Individual livers from each fish were surgically removed and quickly rinsed with cold saline buffer to remove blood. Then, cold methanol was added at 800 μl/100 mg. The whole liver was homogenized in methanol, followed by sonication for 5 min at a 10-W power setting. A total of 30 C. carassius accommodated at 26°C were used in this experiment (n = 10 for each group). The liver from individual C. carassius was treated as one biological sample. Samples were then centrifuged to remove unresolved matter at 12,000 × g, 4°C for 10 min. The supernatant was collected, and 10 μl 0.1 mg/ml ribitol (Sigma-Aldrich, St. Louis, MO, USA) was added as an internal standard. Afterward, the aqueous sample was concentrated in a rotary vacuum centrifuge device (Labconco, Kansas City, MO, USA) for 4 h, and the resultant dried extracts were used for gas chromatography–mass spectrometry (GC-MS) analysis.

The GC-MS analysis was carried out following the two-stage techniques with minor modifications (10). In brief, samples were derivatized in 80 μl of 20 mg/ml methoxamine hydrochloride (Sigma-Aldrich, USA) in pyridine to protect carbonyl moieties through methoximation for 90 min at 37°C, followed by addition of 80 μl of N-methyl-N-trimethylsilyl-trifluoroacetamide (MSTFA) (Sigma-Aldrich, USA) for derivatization of acidic protons at 37°C for another 30 min. The derivatized sample of 1 μl was injected into a 30 m × 250 μm i.d. × 0.25 μm DBSMS column using splitless injection, and analysis was carried out in Agilent 7890A GC equipped with an Agilent 5975 C VL MSD detector from Agilent Technologies (Santa Clara, CA, USA). The initial temperature 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⁻¹, and then held for 5 min. Helium was used as the carrier gas, and the flow rate was kept constantly at 1 ml min⁻¹. 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.

C. carassius (n = 90) were randomly divided into three groups, acclimatized for 7 days at 28°C. Thirty individuals were included in each group. As the fish were of similar weight, each fish was injected with 10 μl 0.85% sterilized saline solution as control group or with 50 and 100 μg fructose as the experimental groups. The fish were injected once daily for 3 days. After the treatment, fish were challenged by intraperitoneal injection of 6.25 × 10⁶ CFU/fish A. hydrophila. Crucian carp were observed for 10 days for accumulative death.

Total RNA of liver was isolated with TRIzol (Invitrogen, Carlsbad, CA, USA). The isolated RNA was then quantified by detecting the intensity of fluorescence. Reverse transcription-PCR was carried out on a Primer-Script™ RT reagent kit with a gDNA eraser (Takara, Japan) with 1 μg of total RNA according to the manufacturer’s instructions. The experiment was performed on six or ten biological samples obtained. qRT-PCR was performed as described previously (47). Primers for each gene are listed in Supplementary Table 1 . Each primer pair was specific. The relative expression of each gene was determined by the comparative threshold cycle method (2^−ΔΔCT method) (48).

After treatment with fructose or sterilized saline solution, fish (n = 10 for each group) were anaesthetized on ice for 10 min to slow the movement of the fish. Blood samples were collected with sterile syringes from the caudal vein around the caudal peduncle into 1.7-ml microcentrifuge tubes without anticoagulants. Collected samples were kept at room temperature for 30 min to allow blood clotting and kept at 4°C overnight. On the second day, blood samples were centrifuged and the supernatant was pipetted into a sterile 1.5-ml microtube. The samples were aliquot and stored at -80°C until use.

For serum lysozyme activity, measurement was performed as was determined by turbidimetric assays as described by Caruso et al. (49) and followed the guidance of Lysozyme Activity Kit (Jiancheng, China). 200 μl of standard lysozyme or serum sample was pipetted into a 5-ml sterilized centrifuge tube and mixed with 2 ml diluted Micrococcus lysodeikticus suspension for less than 10 s to complete the system and measured OD530 in a well-cleaned cuvette. Observation values were taken as T0 and T10 when the cuvette was measured instantly and 10 min later. The lysozyme activity was calculated as the following:

For serum complement activity, alternative complement serum activity was determined as previously described (50, 51). Sheep red blood cells (ShRBC) were washed three times with sterilized pre-cold saline solution and resuspended with ethylene glycol tetraacetic acid–magnesium–gelatin veronal buffer (0.01 M EGTA–Mg–GVB, pH 7). ShRBC suspension was adjusted to 1 × 10⁹ cells/ml. 2 units of ShRBC antibody (Yuanye, China) were used to sensitize prepared ShRBC in a water bath at 37°C with shaking once per 5 min for 30 min. Sensitized ShRBC was washed with EGTA–Mg–GVB buffer two times and resuspended in the same volume of buffer. The complement lysis system including 1.5 ml diluted serum and 1 ml sensitized ShRBC was mixed gently and bathed at 37°C for 30 min. The CH50 standard sample was measured at 0.5 ml 2% ShRBC mixed with 2.5 ml distilled water and incubated at 37°C for 30 min. Then, the sample was centrifuged and the supernatant was collected and measured at OD542. The complement activity was as follows: CH50 (U/mL) = (1/serum volume/mL) * dilution fold.

For bactericidal effect, log-phase A. hydrophila bacterial cultures (OD600 = 0.5) were collected, washed with saline, and resuspended to OD600 = 0.2. 2 ml of the resuspended culture was centrifuged and resuspended in 200 μl serum isolated from either fructose-treated fish or saline-treated fish. The mixture was incubated at 30°C with shaking for 2 h. At last, 1.8 ml saline was added to the system to stop the reaction, which was subsequently used for plating.

To quantify the bacterial loads in the liver, the liver was isolated and collected in a 1.5-ml Eppendorf tube and was kept on ice. The tissue was thoroughly homogenized in 200 μl pre-cold saline buffer (10 mg/ml) by a grinding rod. Then, 100 μl homogenates was serially diluted for plating.

To assess the bactericidal effect of liver homogenates to A. hydrophila, livers from either saline or fructose-treated crucian carp were isolated and homogenized in pre-cold PBS buffer at 5% (m/v). Homogenates were centrifuged at 4°C and 5,000 rpm for 10 min, and the supernatant was collected. 200 μl A. hydrophila (OD600 = 0.2) bacterial culture was mixed with the supernatants at 30°C with shaking. At last, 1.8 ml saline was added to the reaction, which was used for plating by serial dilution.

Crucian carp, Carassius carassius, were infected with serial doses of A. hydrophila (5 × 10⁵, 1 × 10⁶, 2 × 10⁶, 3 × 10⁶, 4 × 10⁶ CFU/fish) by intraperitoneal injection. Fish began to die as early as 24 h postinfection and did not die after 72 h postinfection ( Figure 1A ). The mortality was monitored further for a total of 9 days. Moreover, the dose (3 × 10⁶ CFU) causing 50% of death was selected as LD₅₀, which was used for following analysis.

To understand the metabolic signature that determines the consequences of A. hydrophila infection, C. carassius was challenged with A. hydrophila at LD₅₀. Ten fish were collected for the dying group, survival group, and control group, respectively. The liver—which not only is considered the core organ for the body to maintain physiological functions but also serves as a major organ in teleost (52)—of each fish was individually removed for gas chromatography–mass spectrometry analysis (GC-MS). Therefore, a total of 60 data points were generated. The internal standard, ribitol, as a control to monitor GC-MS performance and known solvent peaks were removed. The same compounds were merged. The Pearson correlation coefficients between two technical replicates were from 0.9991 to 0.9999 ( Supplementary Figure 1A ), indicating that the technical replicates had good reproducibility. A total of 85 metabolites were identified ( Supplementary Figure 1B ), which fell to five functional categories including carbohydrates (52.9%), amino acids (18.8%), lipids (11.8%), nucleotides (7.1%), and others (9.4%) according to Gene Ontology (GO) ( Figure 1B ). The metabolites were summarized and shown as a heatmap ( Figure 1C ). Interestingly, the three groups were separately clustered, suggesting that the dying fish and survival fish mounted differential response, which were also different from the saline control group ( Figure 1C ). In addition, 78 metabolites of differential abundance were identified. This difference was displayed as a Z score plot in Figure 1D , which spanned from -40 to 60 in the dying group including 34 increased metabolites and 44 decreased metabolites and from -20 to 40 in the surviving group including 36 increased metabolites and 42 decreased metabolites. These data together suggest that the metabolome is associated with the state of fish during bacterial infection, highlighting the differential metabolic response in the dying and survival fish.

Metabolic pathway analysis provides the route of chemical change offering a comprehensive view to identify the changes during the biological process (53). Thus, the metabolites of differential abundance were searched against MetaboAnalyst 5.0 (54). A total of 12 metabolic pathways were enriched, and the top eight pathways include alanine, aspartate, and glutamate metabolism, glutamine and glutamate metabolism, β-alanine metabolism, glyoxylate and dicarboxylate metabolism, arginine biosynthesis, aminoacyl-tRNA biosynthesis, citrate cycle, and pyruvate metabolism ( Figure 2A ). The metabolites of each pathway are listed in Figure 2B . For many of the pathways, the metabolites were accumulated during infection despite that several of the amino metabolites were decreased ( Figure 2B ). However, it is also interesting to notice that most of the metabolites of the dying group increased more than those in the survival group ( Figure 2B ). In addition, iPath provides a global view of the metabolic network by integrating different pathways (55). A significant altered part lies on the central carbon metabolism ( Figure 2C ). Correspondingly, the abundance of the metabolites in the citric cycle was decreased in surviving fish. Moreover, the intermediate metabolite of the TCA cycle accumulated higher in the dying group than those in the survival group except citric acid ( Figure 2B ). The pathway enrichment analysis demonstrates that the dying fish is associated with dysregulated central carbon metabolism.

To identify the crucial metabolite(s) that separate the three groups, partial least square discriminant analysis (PLS-DA) was adopted. The three groups were clearly separated from each other ( Figure 3A ), where R2Y = 0.983 > 0.5, indicating good fitting of the model. Discriminating variables were shown as S-plot for both dying vs. control and survival vs. control, where the cutoff values were set as greater or equal to 0.05 for the absolution value of covariance p, and greater or equal to 0.5 for correlation p(corr) ( Figure 3B ). Crucial biomarkers were identified through component p (1) and are represented with red triangles ( Figure 3B ). Thus, a total of 16 crucial biomarkers were identified and displayed as scatter plots ( Supplementary Figure 2 ). Fructose was one of the most outstanding metabolites ( Figure 3C ). Interestingly, the abundance of fructose was decreased in both the dying and surviving groups, but the abundance in the dying group was lower than in the surviving group ( Figure 3C ), implying that fructose might be a survival factor for A. hydrophila infection.

To examine whether fructose can be a survival-promoting factor, C. carassius were injected daily with two concentrations of fructose (50 or 100 μg/fish), which was obtained by scaling up the relative amount to the internal control, ribitol, for 3 days followed by bacterial challenge with A. hydrophila. Under these two concentrations, fructose alone had no effect on the survival of C. carassius ( Figure 4A ). However, upon bacterial challenge, fructose increased the C. carassius survival rate up to 27.2% in a fructose dose-dependent manner ( Figure 4A ), where injection of saline served as control. The relative percent survival rates of the two groups were 37.5% and 56.25% for the groups being treated with 50 and 100 μg/fish, respectively ( Table 1 ). The increased fish survival motivated us to investigate whether this potential was driven by immunity. Being treated with 100 µg fructose the same way as mentioned above, the livers were used for immune gene expression analysis. A total of thirteen genes, tnfα1, tnfα2, ilb1, ilb2, lyz, ifnγ1-1, infγ1-2, nfkbiab, il11, c3, tlr9, tlr2, and tlr3 were included. The expressions of ifnγ1-1, infγ1-2, nfkbiab, tlr3, lyz, and c3 were all increased, but others were slightly decreased ( Figure 4B ). In addition, to confirm if the role of fructose on protecting fish against A. hydrophila infection correlates with immune response, the fish was challenged with a non-lethal dose of bacteria after being pretreated with fructose as in Figure 4B . Consistently, lyz and c3 were the two of the most impacted genes ( Figure 4C ) that increased 4.13- and 2.99-fold, respectively. Thus, these data suggest that fructose influence gene expression that can enhance C. carassius to fight against A. hydrophila infection.

Fructose promotes the expression of innate humoral resonance including lyz and c3, and these two molecules are known for their bacterial lytic activity (56, 57). We postulate that fructose promotes the serum level of lysozyme and complement component 3 that help the elimination of bacteria, which can also be used to confirm our qRT-PCR results. Thus, we measured the serum level of C. carassius after being treated with 100 µg fructose. The lysozyme activity was increased from 64.53 ± 2.31 to 105.48 ± 4.51 U/ml, and the complement activity was increased from 28.55 ± 1.36 to 45.53 ± 3.11 U/ml ( Figures 5A, B ). Moreover, these data corroborate our qRT-PCR data that fructose increased the expression of lysozyme and C3 at both transcriptional level and protein levels. Moreover, we also analyzed whether the increased lysozyme and C3 contribute to the cell killing. Bacteria were incubated with saline, serum, or serum isolated from fructose-treated fish (fructose-serum) for cell killing. As compared to the saline control, the bacteria grew dramatically in the control, whose CFU increased 10.36-fold, whereas the growth in the fructose serum was slowed for 1.9-fold as compared to serum alone ( Figure 5C ). Meanwhile, the liver extracts were also applied. Surprisingly, the growth of the liver extracts from fish pretreated with fructose (fructose-liver extracts) was 5.35-fold slower than the liver extracts alone ( Figure 5D ). These data together suggest that fructose increased the humoral lytic ability toward A. hydrophila.

To support the fructose-mediated bacterial clearance in vivo, we firstly established that the number of bacteria in the survival fish was around 63.31-fold lower than that in the dying fish ( Figure 6A ). The number of viable bacteria was a critical factor for fish survival. C. carassius were treated with either LD50 or a non-lethal dose of bacteria so that the role of fructose in reducing bacterial load can be revealed. Under an LD50 dose challenge, the control group had an increased number of bacteria in 36 h after intraperitoneal infection, which increased 61.44-fold after 36 h postinfection ( Figure 6B ). However, the number in fructose-treated fish was increased less than 9.17-fold at the first 24 h but decreased afterward ( Figure 6B ). Similarly, upon non-lethal bacterial challenge, fructose was also more potent on decreasing bacteria number than by the immune system itself ( Figure 6C ). Thus, these in vitro data demonstrate that fructose enable bacteria clearance that can be further explored in aquaculture.