The overuse or misuse of antibiotics in clinical and agricultural settings has promoted the emergence and spread of antibiotic‐resistant bacteria, some of which are disease‐causing bacterial pathogens (Kraemer et al. , 2019). Meanwhile, resistance to most of the currently used antibiotics now threatens food safety (Wyrsch et al. , 2016). Because diverse mechanisms underlie the antibiotic resistance of different bacterial strains and organisms, it has been challenging to develop novel efficient antibiotics (Lee and Collins, 2011). Under such circumstances, alternative approaches to treat and control infections caused by antibiotic‐resistant pathogens are urgently needed.
The innate immune system, which is conserved throughout evolution, plays a role in activating and initiating the adaptive immune response to bacterial infection (Thomas and Yang, 2016). Because the adaptive immune system is absent in invertebrate species, the role of innate immunity, and its potential use in fighting bacterial infections, can be readily investigated in invertebrates, such as Danio rerio, commonly known as zebrafish (ZF) (Kovarik et al. , 2016; Gammoh and Rink, 2017). Mechanisms of bacterial antibiotic resistance often involve significant genetic changes and may carry a fitness cost, leading to virulence or persistence (Andersson, 2003; Gandon and Vale, 2014), slow growth, compromised pathogenicity and poor transmission (Andersson and Hughes, 2010; Roux et al. , 2015). Antibiotic‐resistant bacterial pathogens also frequently evolve rapidly, adapting to the host environment and avoiding detection and/or destruction by host immune systems. We suggest that the genetic and phenotypic changes that confer antibiotic resistance trigger different host immune responses, and a better understanding of these processes could facilitate efforts to develop effective strategies to manage infections caused by antibiotic‐resistant bacteria.
Metabolism plays an essential role in shaping the innate immune response and strengthening host defence mechanisms (Cheng et al. , 2014; Cameron et al. , 2016; Zhang et al. , 2017; Araldi et al. , 2017). For example, in macrophages, lanosterol decreases expression and activation of signal transducer and activator of transcription 1 (STAT1) and STAT2 by IFN β, which increases survival by reducing cytokine secretion (Araldi et al. , 2017). Besides, under hypoxic conditions, the innate immune response to bacterial infection can be enhanced through the activation of HIF‐1α and neutrophil metabolism (Araldi et al. , 2017). Carbohydrates are also important for modulating immunity (Crocker and Feizi, 1996). For example, protein‐linked oligosaccharides associated with microbes can be recognized by T‐lymphocytes, contributing to immune surveillance (Avci et al. , 2013). Metabolite like L‐aspartic acid promotes the production of nitrogen oxide and enhances phagocytosis to clear bacterial infection (Gong et al. , 2020). But whether metabolism plays roles in distinguishing antibiotic‐sensitive and antibiotic‐resistant bacteria is not yet explored.
Here, we propose that metabolic modulation could be considered as a strategy to boost the effectiveness of the innate immune system (Peng et al. , 2015; Su et al. , 2015). We hypothesize that different host metabolic responses to antibiotic‐sensitive and antibiotic‐resistant bacteria could be exploited to enhance the clearance of bacterial pathogens. In this study, we test this hypothesis using levofloxacin‐resistant (Lev‐R) and levofloxacin‐sensitive (Lev‐S) Vibrio alginolyticus and a zebrafish (ZF) model system. V. alginolyticus causes high morbidity in marine fish, and otitis and wound infection in humans (Newton et al. , 2012; Wang et al. , 2016). Antibiotic‐resistant V. alginolyticus arises frequently (Sperling et al. , 2015), representing a threat to human health and the safe production of farmed fish. Our findings may pave the way for the development of novel alternative strategies for managing antibiotic‐resistant bacterial infections.
In an experimental evolution approach, we generated V. alginolyticus levofloxacin‐resistant (Lev‐R), which was 16MIC of the levofloxacin‐sensitive (Lev‐S) bacterial strain (see Experimental procedures) (Fig. 1A; Fig. S1). Whole‐genome sequencing of Lev‐R V. alginolyticus (Table S1) identified eight mutations, five of which occurred in the following gene coding regions: AT730_01475, aceF, cpxP, cpxA and gyrB. Of these, gyrB is the only gene previously associated with fluoroquinolone resistance (Tables S1 and S2). While cpxP harbours a frameshift mutation, the other four genes harbour point mutations. CpxA and CpxP are bacterial envelope stress proteins previously associated with antibiotic resistance (Danese and Silhavy, 1998; Raivio et al. , 2013; Zhang et al. , 2019). The functions of AT730_01475 and aceF in antibiotic resistance have not yet been elucidated.
Fitness of Lev‐R V. alginolyticus was investigated by injecting 9 × 105 CFU Lev‐R or Lev‐S V. alginolyticus into ZF (Fig. S2). At this dose, infection with Lev‐S V. alginolyticus caused 90% mortality and 10% survival after 2 days, whereas ZF infected with Lev‐R V. alginolyticus had 92% survival 6 days’ post‐injection (Fig. 1B), indicating attenuated pathogenicity of Lev‐R V. alginolyticus. To investigate the adaptation of Lev‐R during infection of ZF, bacterial gyrB gene copies, which were not affected by the presence of point mutations (Fig. S3A), were quantified at different time points after infection. qPCR data indicated 2 × 105 copies of bacterial gyrB immediately after injection (0 h) of ZF with Lev‐S or Lev‐R V. alginolyticus. The rapid proliferation of bacteria was observed from 6 to 24 h post‐injection, with more Lev‐S‐ than Lev‐R V. alginolyticus detected in ZF 24 h post‐injection. However, the number of Lev‐S V. alginolyticus decreased to trace amounts by 48 h and was undetectable at 72 h post‐infection, while the number of Lev‐R V. alginolyticus was similar at 0 and 48 h post‐injection, and was still detectable 72 h post‐injection (Fig. 1C). The bacterial load of injected ZF was also estimated by measuring colony‐forming units (CFUs) in extracts of infected ZF at 0, 6, 24, 48 or 72 h post‐injection. A similar pattern of bacterial growth post‐injection was documented using qPCR‐ or CFU‐based methods to count bacteria (Fig. S4A). Interestingly, bacteria were detected in the liver but not muscle, intestine, spleen or head kidney of infected ZF 72 h post‐infection (Fig. S4B).
To investigate the mechanism that allows Lev‐S V. alginolyticus to be cleared from infected ZF more quickly than Lev‐R V. alginolyticus , spleen (Fig. 1D) and head kidney (Fig. 1E), which are major immunological organs of ZF, were isolated from infected or control ZF, and expression of innate immune genes was analysed by qRT‐PCR, as previously described (Faikoh et al. , 2014). Specifically, expression of 12 genes, Il1b, Il6, il8, il10, il21, tnfa, Ptgr2, ifng, lysozyme, hamp1, hamp2 and leap2, including four antimicrobial peptides, was analysed at 6 h post‐injection. In spleen, Lev‐S and Lev‐R V. alginolyticus induced a stronger immune response than the control, and Lev‐S V. alginolyticus induced a stronger immune response than Lev‐R V. alginolyticus (Fig. 1D). Similar results were obtained in head kidney, except that expression of hamp1 and hamp2 was unaffected (Fig. 1E). However, in head kidney and spleen, no bacteria were detected by qRT‐PCR or plating at this time point (data not shown). Moreover, similar gene expression data were obtained following Lev‐S or Lev‐R V. alginolyticus infection of newborn ZF larvae, whose adaptive immunity is absent so that Lev‐S and Lev‐R induce differential innate immune response can be confirmed (Fig. 1F). Thus, ZF demonstrates slower clearance as well as a weaker immune response to Lev‐R V. alginolyticus; this suggests improved fitness of Lev‐R V. alginolyticus during infection of ZF.
GC‐MS was used to investigate and compare Lev‐R‐ and Lev‐S V. alginolyticus‐induced changes in the ZF metabolome, to identify crucial metabolites as potential biomarkers and to understand the slower clearance of Lev‐R V. alginolyticus from infected ZF. For this purpose, ZF was injected with an LD50 dose of Lev‐S or Lev‐R V. alginolyticus or saline (as a control), and the metabolomes of surviving ZF were analysed by GC‐MS. Ten biological replicates and two technical replicas of each group were performed, yielding 60 data sets. Metabolites that were differentially affected by bacterial infection were identified (Fig. S8). Compared with the control group, 61 or 67 metabolites demonstrated differential abundance in ZF infected with Lev‐S or Lev‐R V. alginolyticus respectively. A Z‐score plot based on the control showed values from −4.60 to 18.26 for metabolites in Lev‐S V. alginolyticus‐infected ZF and from −2.03 to 56.05 for metabolites in Lev‐R V. alginolyticus ‐infected ZF (Fig. 2A). Among the differential abundance of metabolites, 54 metabolites were overlapped, in which 17 were decreased and 23 were increased. In addition, the abundance of 14 metabolites was reversal between Lev‐S and Lev‐R, and the others were Lev‐S‐ and Lev‐R‐specific, including four increased and three decreased in Lev‐S, and eight increased and five decreased in Lev‐R (Fig. 2B). Principal component analysis (PCA) detected obvious separation between the Lev‐S V. alginolyticus‐induced metabolome and Lev‐R V. alginolyticus ‐induced metabolome (Fig. 2C), with no significant outliers. Component t separated Lev‐R V. alginolyticus‐infected ZF from the control group, while component t separated Lev‐S V. alginolyticus‐infected ZF from the other two groups. Discriminating variables were present in the S‐plot using a cut‐off value of ≥ 0.05 for the absolute value of covariance p and a cut‐off value of ≥ 0.5 for p(corr) (Fig. 2D). Crucial metabolite biomarkers were screened by component p and component p. In component p, glucose, maltose, pyroglutamic acid, valine, alanine, tyrosine, threonine, serine, cholesterol, inosine, lactic acid, histidine, hypoxanthine, taurine, phosphoric acid were identified and in component p, maltose, histidine, glycerol 3‐phosphate, GABA, stearic acid, mannose 6‐phosphate, fructose 6‐phosphate, glycine, taurine were identified. Maltose was selected as the most significant biomarker because its abundance was elevated in both component p and component p (Fig. 2D,E), being consistent with our criteria that elevated biomarkers not only distinguished Lev‐R‐induced metabolome from Lev‐S‐induced metabolome, but also separated infectious metabolome, including Lev‐R‐metabolome and Lev‐S‐metabolome, from non‐infected metabolome, the saline control. Indeed, maltose was more abundant in the metabolomes of Lev‐R and Lev‐S V. alginolyticus‐infected ZF than in control, and also more abundant in Lev‐S V. alginolyticus‐infected ZF than in Lev‐R V. alginolyticus ‐infected ZF (Fig. 2F). These results suggest different metabolomic strategies in response to infection by Lev‐S and Lev‐R V. alginolyticus, and that maltose is the most crucial metabolite biomarker.
To explore mechanisms underlying differential abundance of maltose, we measured the transcription of alpha‐amylase, which breaks down starch to form maltose and glucose, as well as maltase‐glucoamylase and alpha‐glucosidase, which metabolize maltose (Fig. 3A). Interestingly, the expression of alpha‐amylase (amy2a) was upregulated in Lev‐S V. alginolyticus‐infected ZF but not in Lev‐R V. alginolyticus‐infected ZF, and the expression of maltase‐glucoamylase (encoded by gaa and ganc) and alpha‐glucosidase (encoded by si) was downregulated to a similar extent in Lev‐S‐ and Lev‐R V. alginolyticus ‐infected ZF (Fig. 3B). More importantly, the promoter of amy2A was induced strongly by Lev‐S‐ but not by Lev‐R V. alginolyticus at 3 h post‐infection (Fig. 3C), after which cell viability decreased and the Lev‐S and Lev‐R induced differential luciferase activity (Fig. S5). Taken together, these results demonstrate that Lev‐S V. alginolyticus‐infected ZF induces higher maltose production through upregulating amy2a transcription to a greater extent than Lev‐R V. alginolyticus. This could imply that the increased abundance of maltose is a metabolic response of the host to V. alginolyticus infection.
Crucial biomarkers can programme the metabolome as a strategy to manage internal or external stressors (Thompson et al. , 2017), which motivated us to examine whether exogenous maltose could potentiate the ability of ZF to kill Lev‐R V. alginolyticus. To examine this possibility, ZF was injected with exogenous maltose. LC‐MS showed that injection of ZF with maltose increased the in vivo maltose level approximately twofold (Fig. 4A) and increased the rate at which bacterial load of Lev‐R V. alginolyticus decreased, being similar to that of Lev‐S V. alginolyticus at 48 and 72 h (Fig. 4B). Similar results were obtained when the bacterial load was quantified using CFUs (Fig. S6). These results indicate that exogenous maltose potentiates the elimination of Lev‐R V. alginolyticus from infected ZF.
We reasoned that the metabolomic change in infected ZF might be tightly associated with the innate immune response to bacterial infection. First, the possibility that maltose affects bacterial growth was examined (Fig. S7A,B), showing that Lev‐R V. alginolyticus grows more slowly than Lev‐S V. alginolyticus, independent of the presence or absence of exogenous maltose. Second, the expression of six V. alginolyticus virulence genes (ctxA, trh, tlh, vpi, toxR and tdh) (Xie et al. , 2005; Abdallah et al. , 2009) was also similar in rich and minimal medium (Fig. S7C,D). However, when transcription of innate immunity genes was quantified in ZF spleen in the presence or absence of exogenous maltose, the results showed >twofold higher expression of lysozyme and the antimicrobial peptides (hamp1 and hamp2) in maltose‐treated ZF than in saline‐treated ZF, no change in expression of il1b, il8, il10 and tnfa, and lower expression of il6, il21, Ptgr2 and ifng (Fig. 5A). In the ZF head kidney, injection of maltose also significantly increased the expression of lysozyme (Fig. 5B).
Previous reports indicate that lysozyme promotes the killing of V. alginolyticus (Sava, 1996; Krens et al. , 2017), prompting us to focus on lysozyme in the following studies. Our evidence shows that maltose increases the activity of lysozyme (Fig. 6A) and that Lev‐S V. alginolyticus is sensitive to killing by lysozyme at 6.25–400 µg ml−1, while Lev‐R V. alginolyticus is less sensitive to killing by lysozyme (e.g. ≥ 100 μg ml−1 lysozyme) (Fig. 6B). Thus, Lev‐R has increased resistance towards dose‐dependent killing by lysozyme.
Lysozyme degrades components of the bacterial cell wall, which promotes clearance of bacteria from the infected host. The binding of lysozyme to the bacterial cell wall involves electrostatic interactions that might be influenced by bacterial membrane potential (Bergers et al. , 1993; Gorbenko et al. , 2007). As the membrane potential is higher for antibiotic‐sensitive bacteria than for antibiotic‐resistant bacteria (Peng et al. , 2015), we hypothesize different binding affinities of lysozyme to Lev‐S and Lev‐R V. alginolyticus. To test this idea, the dose‐dependent binding of lysozyme to Lev‐S and Lev‐R V. alginolyticus was examined at 4°C. The results show relatively lower binding to Lev‐R than to Lev‐S V. alginolyticus at <50 μg ml−1 lysozyme (Fig. 7A). Furthermore, Lev‐R V. alginolyticus demonstrated lower membrane polarization than Lev‐S V. alginolyticus (Fig. 7B), while treatment with 0.125 to 2.0 mM carbonyl cyanide 3‐chlorophenylhydrazone (CCCP) decreased membrane polarization and reduced binding of lysozyme to both strains in a dose‐dependent manner (Fig. 7B,C). Comparatively, lower binding was detected in Lev‐R than in the Lev‐S with lysozyme activity from 7.0 to 4.1 μg ml−1 (reduced 2.9 μg ml−1), and from 8.7 μg ml−1 to 6.3 μg ml−1 (reduced 2.4 μg ml−1 ) respectively (Fig. 7C). These data confirm that lower membrane polarization lowers the binding affinity between bacterial membrane and lysozyme and show that binding affinity also depends on the concentration of lysozyme (Fig. 7D). These results indicate that the lower rate of clearance of Lev‐R V. alginolyticus from ZF can be attributed to lower membrane polarization leading to lower binding of lysozyme to Lev‐R V. alginolyticus.
The present study hypothesizes that the host mounts different innate immune responses to antibiotic‐sensitive and antibiotic‐resistant pathogens, and these responses are associated with different metabolomes. The metabolome can be reprogrammed by exposure to exogenous metabolites, promoting the elimination of antibiotic‐resistant bacteria, which represents a major step in understanding bacterial fitness and its interplay with the innate immune response. Our results show that Lev‐R V. alginolyticus grows slower than Lev‐S (Fig. S7A,B) in vitro and greater survival in vivo in ZF, suggesting associated fitness cost. The fitness cost associated with antibiotic resistance affected the metabolomic strategy of infected ZF and its innate immune response, so that survival of antibiotic‐resistant V. alginolyticus in ZF was greater than the survival of antibiotic‐sensitive V. alginolyticus. There was a close link between the metabolome and the innate immune response, which was demonstrated by discovery metabolomics and the effect of metabolomic reprogramming by exogenous maltose (Peng et al. , 2015). These findings indicate that antibiotic‐resistant V. alginolyticus triggers a weaker innate immune response in ZF than antibiotic‐sensitive V. alginolyticus, which is reflected by their different metabolomes. Thus, metabolomic reprogramming is an effective measure to promote the elimination of antibiotic‐resistant pathogens.
The innate immune system provides immediate defence against bacterial infection (Cameron et al. , 2016); however, infection with Lev‐R V. alginolyticus provoked a weak innate immune response in ZF. Specifically, the expression of lysozyme, Il1b, il6, il8, il10, tnfa, ifng, hamp1, hamp2 and leap2, but not il21, was lower in Lev‐R V. alginolyticus‐infected ZF than in Lev‐S V. alginolyticus‐infected ZF, which is consistent with the improved survival of V. alginolyticus‐infected ZF. These findings explain why antibiotic‐resistant pathogens are less easily eliminated than antibiotic‐sensitive bacteria. However, factors that induce the innate immune response are not yet fully understood.
Reports indicate that the metabolome contributes to host anti‐infection mechanisms (Wang et al. , 2016; Sun et al. , 2019; Xu et al. , 2019), which motivated us to speculate about the role of different metabolomes in coping with antibiotic‐resistant pathogens. To explore this, GC‐MS‐based metabolomics was used to identify reduced maltose as a characteristic feature of the antibiotic‐resistant metabolome as well as reduced expression of innate immunity genes in Lev‐R V. alginolyticus‐infected ZF. Other reports also indicate an association between metabolites and expression of innate immunity (Shapiro et al. , 2014; Correa‐Oliveira et al. , 2016; Haas et al. , 2016; Xu et al. , 2019).
The present study also demonstrates that exogenous maltose increases the expression and activity of lysozyme and antimicrobial peptides. Lysozyme depresses chemotaxis, oxidative metabolism and the generation of superoxide by neutrophils (Sava, 1996), which may explain why maltose elevated expression of lysozyme but reduced expression of other genes. Thus, these results suggest that maltose contributes to the differential expression of innate immune response genes, especially modulating lysozyme expression and activity. This is the first demonstration, to our knowledge, that metabolomic strategy differs in the innate immune response to antibiotic‐resistant and antibiotic‐sensitive bacteria. Alternatively, injection of ZF with maltose may alter the osmolarity of ZF internal fluids (Mittal et al. , 2009), which could alter tissue surface tension and alter the interaction between pathogen and host. This possibility awaits further investigation (Krens et al. , 2017).
Reprogramming the metabolome is a powerful approach to improve the host’s ability to eliminate bacterial pathogens (Thompson et al. , 2017). In the present study, exogenous maltose is used for metabolome reprogramming to restore the ability of ZF to clear antibiotic‐resistant V. alginolyticus. As expected, higher lysozyme expression and activity were detected in ZF injected with maltose. Accordingly, faster elimination of the antibiotic‐resistant bacteria occurred in ZF injected with maltose leading to higher rates of ZF survival. Therefore, maltose plays a crucial role in the elimination of antibiotic‐resistant V. alginolyticus, and this crucial biomarker, identified in the metabolome of Lev‐R V. alginolyticus‐infected ZF, can restore an effective innate immune response to eliminate antibiotic‐resistant pathogens, such as V. alginolyticus.
Evidence indicates that lysozyme is an antibacterial effector of the innate immune system in animals that hydrolyse peptidoglycan (Vanderkelen et al. , 2011). The therapeutic effectiveness of lysozyme is based on its ability to control the growth of bacteria and to modulate host immunity against infections. However, information regarding whether the ability of lysozyme to kill different bacteria is related to its level of expression is not available. We supposed that the ability of lysozyme to kill pathogens depends on the degree of bacterial membrane polarization. To test this idea, we measured the membrane potential of two bacterial strains, Lev‐R and Lev‐S V. alginolyticus, and their binding to lysozyme, showing that they are correlated. Lev‐R V. alginolyticus displayed lower polarization and lower binding than Lev‐S V. alginolyticus. Besides, binding to Lev‐R V. alginolyticus was also related to lysozyme concentration, e.g. when the lysozyme concentration was lower than 50 μg ml−1, the difference was distinct. Furthermore, CCCP, an inhibitor of membrane polarization, decreased the binding of lysozyme to Lev‐R and Lev‐S V. alginolyticus. One possible mechanism is that lysozyme’s isoelectric point is 11.35, such that lysozyme is positively charged, while Lev‐S V. alginolyticus has relatively higher membrane polarization and relatively higher negative charge in binding reaction buffer than Lev‐R V. alginolyticus. This could explain the greater binding affinity of lysozyme to Lev‐S V. alginolyticus. When a large increase in lysozyme concentration is promoted by exogenous maltose, binding to Lev‐R V. alginolyticus increases, causing more efficient Lev‐R V. alginolyticus elimination (Fig. 8). These results support the conclusion that the high membrane polarization is a critical factor for effective lysozyme binding.
Lev‐S V. alginolyticus induces a stronger innate immune response in ZF, including higher expression of cationic peptides like hamp1, hamp2 and leap2, as well as lyz. These molecules are positively charged and prone to bind to the negatively charged bacterial surface. This is consistent with our recent finding that bacterial membrane potential plays a critical role in the recruiting and deposition of complement components on the bacterial surface (Cheng et al. , 2019). Interestingly, exogenous maltose promotes the expression of lysozyme and the elimination of pathogens. These findings demonstrate that the rate of elimination of antibiotic‐resistant bacteria may depend on lysozyme abundance and bacterial membrane polarization. Furthermore, these findings also indicate that metabolic modulation can promote the innate immune response and clearance of antibiotic‐resistant bacteria.
The current study does not identify the mechanism by which maltose regulates lysozyme expression. Maltose can be transported into the cell and degraded to glucose, where it enters the glycolysis pathway. Dietary sugars, including starch, dextrin, maltose, glucose and cellulose, could potentially increase lysozyme level and activity in ZF serum (Kong et al. , 2019). However, maltose is more potent than glucose in elevating lysozyme expression. This is consistent with our data that maltose is more abundant in Lev‐S V. alginolyticus‐infected than in Lev‐R V. alginolyticus‐infected ZF, with limited glucose production, highlighting the role of maltose in regulating lysozyme expression.
The regulation of lysozyme transcription has been extensively studied in chicken. Upon stimulation with LPS or other pro‐inflammatory molecules, the transcription of lysozyme is rapidly activated by transcription factors NF1 and Fli‐1, which displace CTCF (Kontaraki et al. , 2000; Lefevre et al. , 2003; Lefevre et al. , 2008). However, the relationship between sugar metabolism and NF1 and Fli‐1‐dependent regulation of lysozyme expression has not yet been investigated.
The last thing worthy of mention here is the role of antimicrobial peptides in killing antibiotic‐resistant bacteria. As shown in Figs 2D and 6C, the expression of hamp1, hamp2 and lysozyme is similar. hamp1 and hamp2 encode hepcidin, an antimicrobial peptide highly conserved in vertebrates (Shike et al. , 2004). Hepcidin also has roles in iron homeostasis in macrophages and is induced under stress (Li et al. , 2019). Thus, it is worth exploring this role further.
In summary, the present study indicates that a host can mount different metabolic strategies in response to infection with antibiotic‐sensitive and antibiotic‐resistant bacteria and that there is a close relationship between the metabolomic strategy and the innate immune response. Crucial biomarkers associated with the metabolomic strategy lead to differences in the innate immune response, which can be modulated by these biomarkers. These findings highlight the role and potential of metabolomic strategies to promote the innate immune response to antibiotic‐resistant pathogens.
This study was conducted following 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 (http://ZFIN.org). All experiments were approved by the Institutional Animal Care and Use Committee of Sun Yat‐sen University (Animal welfare Assurance Number: 16).
The bacterial strain used in this experiment is V. alginolyticus 12G01, a collection of our laboratory. To generate levofloxacin‐resistant V. alginolyticus (Lev‐R), V. alginolyticus was subcultured in LB medium with 1/2 minimum inhibitory concentration (MIC) of levofloxacin (0.125 µg) for serial passages, which led to the generation of 16 MIC levofloxacin‐resistant V. alginolyticus. The V. alginolyticus cultured in the same way but in the absence of levofloxacin is named as Lev‐S. MIC was determined by antimicrobial susceptibility testing as previously described (Peng et al. , 2015). The overnight culture was subcultured in fresh LB medium and grew until the optical density was 0.5 (OD 600 nm) at 30°C. The log‐phase cells were diluted and dispensed into 96‐well microtitre polystyrene tray with 105 cells each well. The tray contained a series of twofold dilutions of antibiotics. After 16 h of incubation, the MIC was defined as the lowest concentration that completely inhibits growth as determined OD at 600 nm. Three biological repeats with two technical replicates were carried out for each sample.
Zebrafish, 45 days post‐fertilization (dpf) (0.20 ± 0.03 g in weight and 2.50 ± 0.24 cm in length), were obtained from a zebrafish breeding Corporation, Guangzhou, P.R. China. These animals were free of Vibrio species infection through microbiological detection and reared in 25‐l open circuit filtered water tanks at room temperature with aeration. They were maintained in our laboratory for another two weeks before experimental manipulation. Zebrafish were fed with regular food Pellet Feed, Shandong Weifang Yeebrand Food Co., LTD, in a maintenance ration so a small and steady growth was expected on a 12/12 h rhythm of light and darkness photoperiod always.
Three‐day newborn larvae were obtained as previously described (Winter et al. , 2008). Fish were mated in discrete male–female pairs to allow individual clutch use. When sufficient embryos were deposited, they were subjected to an antifungal ‘bleaching’ procedure: 1% (W/V) chloramine T (Sigma‐Aldrich, St Louis, MO, USA) in culture water for 1 min, then two rinces in culture water followed by transfer to a Petri dish containing 0.8 µm sterile‐filtered culture water. Throughout the culture period, a Petri dishwater change was undertaken every 24 h to maintain water quality.
A total of 660 D. reiro were injected with saline (n = 60), Lev‐S (n = 300) or Lev‐R (n = 300), where the bacterial dose for Lev‐S and Lev‐R is 9 × 105 CFU per fish. The survival fish were collected and proceed for sample preparation for GC‐MS as previously described (Xu et al. , 2019). D. rerio, 60 dpf, 0.22 ± 0.04g in weight and 2.52 ± 0.31 cm in length, were euthanized in ice slush (5 parts ice/1 part water, 0–4°C) for at least 10 min following cessation of gill movement and left in the ice water for a total of 20 min after cessation of all movement to ensure death by hypoxia following the guidelines of NIH (Reed and Jennings, 2011). D. rerio were rinsed with distilled water and then wiped thoroughly with sterilized gauze. Each fish was cut into five pieces on ice so the metabolites of the body fluids will diffuse out and then weighted. The appropriate volume of saline (100 μl 100 mg−1) was added according to the weight. After centrifugation at 3000 g at 4°C, 100 μl fluid was aliquoted for the following metabolomic analysis. Metabolites were extracted with 0.2 ml of cold methanol, containing 10 μl of 0.1 mg ml−1 ribitol (Sigma‐Aldrich) as an analytical internal standard. After centrifugation at 12 000 g for 10 min, the supernatant was concentrated in a rotary vacuum centrifuge device, LABCONCO. The dried polar extracts were used for GC‐MS analysis.
The GC‐MS analysis was carried out with a variation on the two‐stage techniques as described previously (Su et al. , 2018). In brief, samples were derivatized in 40 μl of 20 mg ml−1 methoxyamine hydrochloride (Sigma‐Aldrich) in pyridine to protect carbonyl moieties through methoximation for 90 min at 37°C, followed by adding 80 μl of N‐methyl‐N‐trimethylsilyl trifluoroacetamide (MSTFA) for derivatization of acidic protons (Sigma‐Aldrich) 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 DBS‐MS column using splitless injection, and analysis was carried out in Agilent 7890A GC equipped with an Agilent 5975C VL MSD detector, Agilent Technologies. 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 then held for 5 min. Helium was used as the carrier gas and flow was kept constantly at 1 ml min−1. 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.
Metabolites were identified following the guidelines of the metabolomics standards initiative (Sumner et al. , 2007). In conclusion, the deconvolution and calibration of the acquired mass spectra were performed with AMDIS (Agilent Open LAB CDS ChemiStation C.01.01). To avoid false positives, peaks with a signal‐to‐noise ratio (S/N) lower than 30 were excluded (Pezzatti et al. , 2019). Additionally, the artefact peaks were removed through comparison with the blank samples. Metabolites were identified by retrieving their mass spectra in the NIST 2011 (National Institute of Standards and Technology, Gaithersburg, MD, USA) library and GMD 2011 (Golm Metabolome Database, Potsdam, Germany) according to the following criteria: match value ≥ 750, reverse match value ≥ 800 and a probability ≥ 60% (Zhu et al. , 2016). The relative peak area value of the ribitol was taken as the internal standard to calculate the metabolite abundance.
D. rerio were randomly divided into control and test groups and acclimatized for 14 days at 28°C after purchase from the fish corporation. Thirty individuals were included in each group. D. rerio were anesthetized by immersing in 0.02% tricaine methanesulfonate solution, which was subsequently injected with 225 μg maltose, a dose that was estimated from the average concentration of maltose after Lev‐S infection (225 μg per fish), into individual D. rerio as test group or the same volume of sterile saline as control group. Both groups were injected once daily for 3 days. After the last shot, D. rerio were challenged by intramuscular injection of 1.0 × 106 CFU of Lev‐S/fish or 3.0 × 106 CFU of Lev‐R/fish. These zebrafish were observed for symptoms twice daily for 15 days for accumulative death.
Bacteria were grown in 100 ml LB medium at 30°C overnight and then diluted 1:100 into fresh LB medium until its absorbance value of OD600 equal to 0.2. Then, the cells were washed with a saline buffer three times and were resuspended in saline buffer. D. rerio were divided into two groups treated with and without maltose. The group with maltose was injected with 225 μg maltose with a microlitre syringe (scale from 0.00 to 0.05 ml), Shanghai Gaoge Industrial and Trading Co.LTD, China, as test group and the group without maltose were injected with the same volume of sterile saline as a control group for three days. Samples were collected for lysozyme activity assay by commercial kit and maltose quantification by UPLC‐MS. Meanwhile, each of the other fish from the two groups was challenged with 2 × 105 CFU of Lev‐R by intramuscular injection. For qPCR detection of bacterial counts, D. rerio at 0, 6, 24, 48, 72 h post‐injection were euthanized in ice slush (5 parts ice/1 part water, 0–4°C) for at least 10 min following cessation of gill movement, and left in the ice water for a total of 20 min after cessation of all movement to ensure death by hypoxia following the guidelines of NIH (Reed and Jennings, 20112011). Total DNA was extracted from samples with DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions. The purified DNA was stored at − 20°C for qPCR amplification. The experiment was performed in six biological replicates. For qRT‐PCR detection of innate immune genes, D. rerio at 6 h post‐injection was anesthetized in 0.02% tricaine methanesulfonate solution. Spleens and head kidneys were removed ascetically and were stored at −80°C until RNA extraction. Spleens or head kidneys from three D. rerio were pooled as one sample for RNA isolation. To perform qRT‐PCR in D. rerio newborn larvae, larvae of 3 days dpf were collected and immersed in 1 × 105 V. alginolyticus in Hank’s buffer for three days. After infection, larvae were immediately collected and three larvae were pooled as one sample for RNA isolation.
Total RNA was isolated with Trizol (Invitrogen, Carlsbad, CA, USA) and RNA was then quantified by NanoDrop, Thermo Scientific. Reverse transcription‐PCR was carried out on a Prime‐ScriptTM RT reagent Kit with gDNA eraser (Takara, Kusatsu, Shiga, Japan) with 1 μg of total RNA according to manufacturer’s instructions. The experiment was performed in six biological replicates.
Membrane polarization was carried out as described previously (Su et al. , 2018). DiOC2(3), a fluorescent membrane potential indicator dye, was applied to investigate the bacterial membrane potential (membrane polarization). Bacteria were obtained from log‐phase cultures and diluted to approximately 1 × 106 cells per ml in filtered saline. These bacteria were treated with or without 1 mM CCCP for 30 min. Then, a volume of 10 µl of 3 mM DiOC2(3) was added and then incubated at room temperature for 30 min. Stained bacteria were assayed with a flow cytometer equipped with a laser emitting at 488 nm. Fluorescence was measured on the green and red channels (‘GC’ and ‘RC’), and filters were used to detect fluorescein and Texas Red dye respectively. Because the relative amount of red and green fluorescence intensity varied with cell size and aggregation, the ratio of red to green fluorescence intensity was used as a size‐independent indicator of membrane potential.
Bacteria were grown in 100 ml LB medium at 30°C overnight. Bacteria were washed three times with saline buffer and were resuspended in saline to OD 600 at 1.0. A doubling dilution series ranging from 6.25 to 400 µg ml−1 of lysozyme, Sangon Biotech, Shanghai, China, were used for quantitative analysis of lysozyme captured by Lev‐R or Lev‐S. The resuspended cells were mixed with 200 µl lysozyme at 4°C for 16 h. For the experiment with carbonyl cyanide 3‐chlorophenylhydrazone (CCCP), bacteria were pretreated with different CCCP for 30 min before adding lysozyme. After incubation, the mixture was washed with cold saline, followed by washing with 200 µl 0.1 M Gly‐HCl buffer (pH 2.4). Finally, the mixture was centrifuged at 7000 g for 15 min at 4°C, the supernatant was collected and neutralized with 1 M Tris–HCl (pH 8.8), and stored at −20°C until lysozyme activity assay.
Lysozyme Assay Kit was purchased from the Nanjing Jiancheng Bioengineering Institute, China. The activity of lysozyme in zebrafish body fluid and captured by Lev‐S and Lev‐R was determined according to the method based on the lysis of the lysozyme‐sensitive Gram‐positive bacteria Micrococcus lysodeikticus. Shortly, 2 ml of M. lysodeikticus (g‐type) lysozyme (Sigma‐Aldrich) at a concentration of 0.2 mg ml−1 (w/v) in 50 mM phosphate buffer (pH 6.2) and 100 µl body fluid or reaction mixture was added. The reduction in absorbance at 530 nm was measured after 15 min water bath at 37°C. Results were expressed in units of lysozyme ml−1. One unit was defined as the amount of sample causing a decrease in absorbance of 0.001 units per min.
A single bacterial colony was grown in 50 ml LB broth in 250‐ml flasks for 24 h at 30°C. After centrifugation at 8000 rpm for 5 min, samples were washed twice with 30 ml sterile saline and resuspended in M9 minimal media supplemented with 10 mM acetate, 1 mM MgSO4 and 100 mM CaCl2 to 0.2 at OD600. Lysozyme (g‐type) or lysozyme plus serum was added and incubated at 30°C for 1 h. After incubation, 100 ml aliquot samples were periodically removed, serially diluted and plated (10 ml aliquots) onto LB agar plates. The plates were cultured at 30°C for 8 h. Only those dilutions yielding 20–200 colonies were enumerated to calculate colony‐forming units (CFUs). Per cent survival was determined by dividing the CFU obtained from a treated sample by the CFU obtained from control.
Danio rerio cell line, ZF4, was purchased from the Institute of hydrobiology, China, and was maintained in DMEM: F12 medium, GIBICO, Life Technologies, supplemented with 10% foetal bovine serum. Cells were grown in a humidified incubator at 5% CO2 and 28°C.
To perform promoter activity assay, the 1500 bp upstream of the exon 1 (primers were shown in Table S3) were amplified by PCR and cloned into a promoterless vector, pGL3.0‐Basic, upstream of the firefly luciferase gene. The insert was confirmed by sequencing. ZF4 cells were harvested using 0.25% (W/V) trypsin solution and seed into 24‐well plates at a density of 2.5 × 105 cells per well. The plasmid was transfected into the cell with polyethyleneimine. To minimize non‐relevant influences, an additional vector (pRL‐TK) containing the Renilla luciferase gene was co‐transfected. After an incubation time of 24 h, the cells were co‐cultured with 1 × 106 Lev‐s or Lev‐R for 3 h. Then, the cells were lysed and incubated with luciferin according to the manufacturer’s instructions. The luciferase activity of the promoter‐firefly luciferase constructs was then normalized to the Renilla luciferase activity derived from the pRL‐TK vector and expressed as relative fold.
Data shown are the means ± SEM. The Gaussian distribution of data was analysed by D’Agostino‐Pearson omnibus normality test (Prism 7.0, GraphPad Software, San Diego, CA, USA) and Kolmogorov–Smirnov test (Prism 7.0). The variance of data was analysed by the homogeneity of variance test (SPSS 22.0, IBM Corp, Armonk, NY, USA) or Brown–Forsythe test (Prism 7.0).
As none of the dependent variables comply with the ANOVA assumptions of normality, therefore, data for one‐factorial design were analysed by Kruskal–Wallis followed by Dunn’s multiple comparison post hoc test. (Prism 7.0). Data for 2‐factorial designs were analysed by Aligned Rank Transform procedure with ANOVA (ART‐ANOVA), and ART‐ANOVA is a non‐parametric factorial analysis technique recently introduced by Wobbrock et al. (2011), which enables the use of ANOVA test after alignment and ranking and maintains the integrity of interaction effects of factors. Statistical details of all experiments can be found in the figure legends, and significance is described in the figure legends as ∗P < 0.05, ∗∗P < 0.01.
The multivariate analysis was performed using principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS‐DA) with SIMCA 12.0 software (Umetrics, Umeå, Sweden). PCA was used to describe the sample variations between groups (Saline, Lev‐S, and Lev‐R) in the scoring matrix. OPLS‐DA was used to assign samples according to the classes, variability in the S‐Plot. The potential biomarkers were selected when p(corr) > 0.5 combined with P ≥ 0.05.