REGULATION OF HEMOLYSIN GENE EXPRESSION AND THE EFFECTS OF METABOLISM MUTATIONS ON THE VIRULENCE OF VIBRIO ANGUILLARUM

Vibrio anguillarum is the causative agent of vibriosis, a fatal hemor rhagic septicemic disease. V. anguillarum infects more than 50 fresh and salt-water fish species including various species of economic impor tance to the larviculture and aquaculture industry. In vibriosis, V. anguillarum invades its host fish through the intestine and skin. Infected fish usually die with systemic infection of V. anguillarum. Bacterial hemolysins are exotoxins that cause lysis of erythrocytes in the host and thus the release of the intracellular heme, and are ther efore identified as important virulence factors. Moreover, hemolysins are known t o also cause lysis in other cell types, including mast cells, neutrophils and polymo rph nuclear cells. Two hemolysin gene clusters, vah1-plp and rtxACHBDE, have been previously identified and described. The activities of the protein encoded by the plp gene were not known. In the first manuscript, we describe the biochemica l activities of the plp-encoded protein and its role in pathogenesis. The plp gene, one of the components in vah1 cluster, encodes a 416-amino-acid protein (Plp), w hich has homology to lipolytic enzymes containing the catalytic site ami no acid signature SGNH. Hemolytic activity of the plp mutant increased 2-3-fold on sheep blood agar indi cating that plp represses vah1; however, hemolytic activity of the plp mutant decreased by 2-3-fold on fish blood agar suggesting that Plp has different e ffects against erythrocytes from different species. His 6-tagged recombinant Plp protein (rPlp) was over-exp r ssed in E. coli. Purified and re-folded active rPlp exhibited phos pholipase A2 activity against phosphatidylcholine and no activity against phospha tidylserine, phosphatidylethanolamine, or sphingomyelin. Charact erization of rPlp revealed broad optimal activities at pH 5–9 and at temperatures of 30-64°C. Divalent cations and metal chelators did not affect activity of rPlp. We also demonstrated that Plp was secreted using thin layer chromatography and immunoblot anal ysis. Additionally, rPlp had strong hemolytic activity towards rainbow trout ery throcytes, but not to sheep erythrocytes suggesting that rPlp is optimized for lysis of phosphatidylcholine-rich fish erythrocytes. Further, only the loss of the plp gene had a significant effect on hemolytic activity of culture supernatant on fish erythrocyte s, while the loss of rtxA and/or vah1 had little effect. However, V. anguillarum strains with mutations in plp or in plp and vah1 exhibited no significant reduction in virulence co mpared to the wild type strain when used to infect rainbow trout. In the next manuscript, we used degenerate PCR to i dentify a positive hemolysin regulatory gene, hlyU, from the unsequenced V. anguillarum genome. The hlyU gene of V. anguillarum encodes a 92-amino acid protein and is highly homologous to other bacterial HlyU proteins. An hlyU mutant was constructed, which exhibited ~5-fold decrease in hemolytic activi ty on sheep blood agar with no statistically significant decrease in cytotoxicity of the wild type strain. Complementation of the hlyU mutation restored both hemolytic and cytotoxic act ivity. Both semi-quantitative RT-PCR and real time RT-qPCR were used to examine expression of the hemolysin genes under exponential and stationary phase conditions in wild type, hlyU mutant, and hlyU complemented strains. Compared to the wild type strain, expression of rtx genes deceased in the lyU mutant while expression of vah1 and plp was not affected in the hlyU mutant. Complementation of the hlyU mutation restored expression of the rtx genes and increased vah1 and plp expression to levels higher than in the wild type. The transcript ional start sites in the intergenic regions of both vah1/plp and rtxH/rtxB genes were determined using 5’-RACE and the binding sites for purified HlyU was discovered usin g DNA gel mobility shift experiments and DNase protection assays. In the third manuscript, we identified the ns gene, which encodes the H-NS protein, and acts as a negative regulator of both g ene clusters. The V. anguillarum H-NS protein shares strong homology with other bact erial H-NS proteins. An hns mutant exhibited increased hemolytic activity and c ytotoxicity compared to the wild type strain. Complementation of the ns mutation restored hemolytic activity and cytotoxicity levels to near wild type levels. Furt her, expression of rtxA, rtxH, rtxB, vah1 and plp increased in the hns mutant, and decreased in the complemented hns mutant strain when compared to the wild type strain . Additionally, experiments using DNase I, showed that purified recombinant H-N S protected multiple sites in the promoter region of both gene clusters. The hns mutant also exhibited significantly attenuated virulence against rainbow trout. Comple entation of the hns mutation restored virulence to wild type levels, suggesting hat H-NS regulates many genes that affect fitness and virulence. Previously, we showe d that HlyU is a positive regulator of expression for both gene clusters. In this study , we demonstrate that up-regulation by hlyU is hns-dependent, suggesting that H-NS acts to repress or ilence both gene clusters, and HlyU acts to relieve that repression or silencing. In the last manuscript, we aimed to create avirulen t and immunogenic V. anguillarum strains that can be used as a live vaccine for fis h, without knocking out any of the hemolysin genes. For this purpose, six g enes ( mdh, icd, sucA, sucC, sdhC, and fumA) encoding enzymes in tricarboxylic acid (TCA) cycl e and one gene ( cra) encoding a fructose metabolism repressor were ident ified and mutated in wild type V. anguillarum M93Sm (serotype O2a). Among all mutants, icd mutant showed high attenuation of virulence and lowest cell density li mit in two forms of rich media. All mutants exhibited the same or higher levels of hemo lysin gene expression compared to wild type during log phase. Further, fish that were p -treated by immersion with icd mutant protected rainbow trout from the subsequent challenge of V. anguillarum M93Sm; and fish that were pre-treated with injectio n f the icd mutant elicited cross-serotype immunity against the subsequent chal lenge of V. anguillarum NB10 (serotype O1). The results suggest the TCA cycle mu tation approach is likely to be an easy method to construct modified live vaccine for a wide variety of pathogenic bacteria.

The activities of the protein encoded by the plp gene were not known.
In the first manuscript, we describe the biochemical activities of the plp-encoded protein and its role in pathogenesis. The plp gene, one of the components in vah1 cluster, encodes a 416-amino-acid protein (Plp), which has homology to lipolytic enzymes containing the catalytic site amino acid signature SGNH. Hemolytic activity of the plp mutant increased 2-3-fold on sheep blood agar indicating that plp represses vah1; however, hemolytic activity of the plp mutant decreased by 2-3-fold on fish blood agar suggesting that Plp has different effects against erythrocytes from different species. His 6 -tagged recombinant Plp protein (rPlp) was over-expressed in E.
coli. Purified and re-folded active rPlp exhibited phospholipase A2 activity against phosphatidylcholine and no activity against phosphatidylserine, phosphatidylethanolamine, or sphingomyelin. Characterization of rPlp revealed broad optimal activities at pH 5-9 and at temperatures of 30-64°C. Divalent cations and metal chelators did not affect activity of rPlp. We also demonstrated that Plp was secreted using thin layer chromatography and immunoblot analysis. Additionally, rPlp had strong hemolytic activity towards rainbow trout erythrocytes, but not to sheep erythrocytes suggesting that rPlp is optimized for lysis of phosphatidylcholine-rich fish erythrocytes. Further, only the loss of the plp gene had a significant effect on hemolytic activity of culture supernatant on fish erythrocytes, while the loss of rtxA and/or vah1 had little effect. However, V. anguillarum strains with mutations in plp or in plp and vah1 exhibited no significant reduction in virulence compared to the wild type strain when used to infect rainbow trout.
In the next manuscript, we used degenerate PCR to identify a positive hemolysin regulatory gene, hlyU, from the unsequenced V. anguillarum genome.
The hlyU gene of V. anguillarum encodes a 92-amino acid protein and is highly homologous to other bacterial HlyU proteins. An hlyU mutant was constructed, which exhibited ~5-fold decrease in hemolytic activity on sheep blood agar with no statistically significant decrease in cytotoxicity of the wild type strain.
Complementation of the hlyU mutation restored both hemolytic and cytotoxic activity. using DNase I, showed that purified recombinant H-NS protected multiple sites in the promoter region of both gene clusters. The hns mutant also exhibited significantly attenuated virulence against rainbow trout. Complementation of the hns mutation restored virulence to wild type levels, suggesting that H-NS regulates many genes that affect fitness and virulence. Previously, we showed that HlyU is a positive regulator of expression for both gene clusters. In this study, we demonstrate that up-regulation by hlyU is hns-dependent, suggesting that H-NS acts to repress or silence both gene clusters, and HlyU acts to relieve that repression or silencing.
In the last manuscript, we aimed to create avirulent and immunogenic V.
anguillarum strains that can be used as a live vaccine for fish, without knocking out any of the hemolysin genes. For this purpose, six genes (mdh, icd, sucA, sucC, sdhC, and fumA) encoding enzymes in tricarboxylic acid (TCA) cycle and one gene (cra) encoding a fructose metabolism repressor were identified and mutated in wild type V.
anguillarum M93Sm (serotype O2a). Among all mutants, icd mutant showed high attenuation of virulence and lowest cell density limit in two forms of rich media. All mutants exhibited the same or higher levels of hemolysin gene expression compared to wild type during log phase. Further, fish that were pre-treated by immersion with icd mutant protected rainbow trout from the subsequent challenge of V. anguillarum M93Sm; and fish that were pre-treated with injection of the icd mutant elicited cross-serotype immunity against the subsequent challenge of V. anguillarum NB10 (serotype O1). The results suggest the TCA cycle mutation approach is likely to be an
Several genes have been reported to be correlated with virulence by V. anguillarum, including the vah1 hemolysin cluster [7,8], the rtx hemolysin cluster [9], the siderophore mediated iron transport system [10], the empA metalloprotease [2,5], and the flaA gene [11]. Hemolytic activity of V. anguillarum has been considered to be the virulence factor responsible for hemorrhagic septicemia during infection [10]. We have identified two hemolysin gene clusters in V. anguillarum that contribute to the virulence of this pathogen [8,9]. One gene cluster, rtxACHBDE, encodes a MARTX toxin and its type I secretion system [9]. The second hemolysin gene cluster in V.
anguillarum strain M93Sm contains the hemolysin gene vah1 linked to two putative lipase-related genes (llpA and llpB) and a divergently transcribed hemolysin-like gene (plp) that appears to function as a repressor of hemolytic activity [8]. The plp-encoded protein has very high sequence similarity to phospholipases found in other pathogenic Vibrio species [8]. However, the enzymatic characteristics of Plp in V. anguillarum were not described.
Generally, phospholipases are divided into several subgroups depending on their specificity for hydrolysis of ester bonds at different locations in the phospholipid molecule. Phospholipases A (PLAs) cleave long chain fatty acids at sn-1 (PLA1) or sn-2 (PLA2) position from phospholipid to yield lysophospholipid and free fatty acid; phospholipases C (PLCs) cleave phospholipid into diacylglycerol and a phosphate-containing head group; and phospholipases D (PLDs) cleave phospholipid into phosphatidic acid and an alcohol. It is known that some phospholipid products are used as secondary messages, which play a central role in signal transduction [12].
In this study, we determined that plp encodes a phospholipase A2 in V. anguillarum, and then purified recombinant Plp protein (rPlp) from E. coli to investigate biochemical characteristics. We also described the contribution and specificity of rPlp for hydrolysis of phospholipids, and its ability to lyse fish erythrocytes.

Results
Identification of a putative phospholipase gene plp. Previously, a putative phospholipase gene, plp, was identified in the vah1 hemolysin cluster of V.
anguillarum strain M93Sm [8]. The 1251-bp plp gene (Genbank accession EU650390) was predicted to encode a protein of 416 amino acids. A BLASTx [13]  and rtxA resulted in a hemolysis negative mutant when plated on TSA-sheep blood agar [9]. Similar results were observed when using Luria-Bertani broth plus 2% NaCl plus 5% sheep blood (LB20-sheep blood) agar (data not shown). However, on LB20 plus 5% rainbow trout blood (LB20-rainbow trout blood) agar, the plp mutant exhibited a smaller zone of hemolysis compared to wild type strain M93Sm (Fig. 2B); complementation of plp restored the hemolytic activity of the mutant strain (Fig. 2B).
Similar results were observed when using LB20 plus 5% Atlantic salmon blood agar (data not shown), suggesting that the ability of Plp to lyse erythrocytes is dependent upon the source of erythrocytes and, therefore, their lipid composition.  Fig. 3 and 6A). However, rPlp had almost no activity against both NBD-phosphatidylethanolamine (NBD-PE) (Fig. 6B) and NBD-phosphatidylserine (NBD-PS) (Fig. 6C), showing only 2% and 5%, respectively, of the activity of the standard PLA2 protein against each of the substrates. The data indicated that the rPlp protein does not efficiently cleave either phosphatidylethanolamine or phosphatidylserine. Additionally, unlike the standard sphingomyelinase (Sigma), rPlp was not able to cleave the NBD-sphingomyelin into the NBD-ceramide and phosphocholine (Fig. 6D), indicating that rPlp had no sphingomyelinase activity.
Taken together, the data demonstrated that Plp is a phosphatidylcholine-specific PLA2 enzyme.
rPlp is able to lyse the fish erythrocytes directly. Membrane phospholipid compositions are quite varied among the animal species, especially for phosphatidylcholine. It is known that phosphatidylcholine makes up 58% of the total phospholipid in fish erythrocytes [18]; however, no phosphatidylcholine is found in sheep erythrocytes [19]. In order to determine whether the differential hemolysis observed for plp mutants of V. anguillarum (Fig. 2) is due to the activity of Plp against PC, we tested the ability of purified rPlp to lyse Atlantic salmon erythrocytes.
Addition of recombinant Plp resulted in the lysis of Atlantic salmon erythrocytes, with the amount of lysis directly related to the amount of rPlp added to the blood suspension ( Fig. 7). In contrast, addition of rPlp to a suspension of sheep erythrocytes resulted in no lysis of those cells (Fig. 7). These data show that Plp has phosphatidylcholine-specific phospholipase A2 activity and can directly lyse fish erythrocytes.

Plp is one of the hemolysins of V. anguillarum.
Previously, we demonstrated that there are two major hemolysin gene clusters in the M93Sm, the vah1 cluster [8] and the rtxA cluster [9]. Mutation of both vah1 and rtxA completely eliminated the hemolytic activity of M93Sm on TSA-sheep blood agar [9]. While, mutation of the plp gene resulted in 2-3-fold increased hemolytic activity on TSA-sheep blood agar, with vah1 expression increased both transcriptionally and translationally in the plp mutant, indicating that Plp is a putative repressor of vah1 [9], Plp also has hemolytic activity against fish erythrocytes due to its phosphatidylcholine-specific activity (Figs. 6 and 7). To investigate the relationship of the three hemolysins, culture supernatants obtained from various V. anguillarum strains (Table 1) were used to examine the hemolytic activity against the fish blood ( Table 2).
In contrast to the strong hemolytic activity against 5% rainbow trout blood mixed with culture supernatant from the wild type strain M93Sm, hemolytic activity of culture supernatant from strain S262 (plp) declined by >70% (Table 2). Additionally, all mutants containing a knockout of plp exhibited significant declines (P< 0.05) in hemolytic activity. The triple hemolysin mutant, XM90 (plp vah1 rtxA) had no ability to lyse fish erythrocytes (Table 2). However, mutations in either vah1 or rtxA, but not plp, resulted in little or no decline in hemolytic activity against fish erythrocytes compared to supernatants from wild type cells (Table 2). Further, complementation of plp restored the hemolytic activity of supernatants from both the plp-complemented strains (XM31, plp+ and XM93, vah1 rtxA plp+) ( Table 2).Taken together, these data clearly demonstrate that Plp is the major hemolytic enzyme responsible for the lysis of fish erythrocytes by culture supernatants of V. anguillarum.
Plp is not a major virulence factor for V. anguillarum during fish infection. In order to determine whether the plp gene affects virulence in fish, an infection study was performed by inoculating rainbow trout by IP injection with either the wild type strain M93Sm or mutant strains S262 (plp) or JR03 (vah1plp). The results of this experiment (Fig. 8) indicated that there were no statistical differences in mortality between the three strains. This suggested that mutation of either or both plp or vah1 did not decrease the virulence of M93Sm. These results are consistent with our previous observations that rtxA is a major virulence factor of M93sm and that mutation of vah1 does not affect virulence [8], and demonstrate that Plp is not a major virulence factor in the V. anguillarum M93Sm.
In this report, we describe the characteristics of the V. anguillarum phospholipase protein (Plp) encoded by plp, and its contribution to the hemolytic activity of V.
anguillarum. Specifically, we show that Plp is a secreted phospholipase with A2 activity with specificity for phosphatidylcholine. The enzyme has a broad temperature optimum (37 -64°C) and a broad pH optimum (pH5.5 -8.7).
Phospholipases are broadly distributed among the Vibrionaceae and often contribute to the virulence of the pathogenic members of this family. For example, the TLH (synonym: lecithin-dependent hemolysin, LDH) of V. parahaemolyticus [20][21][22] was the first well-studied lecithin-dependent PLA / lysophospholipase [23]. A lecithinase (encoded by lec) was also identified in V. cholerae [24]. Fiore et al [24] found that a lec mutant strain was unable to degrade lecithin and the culture supernatant exhibited decreased cytotoxicity. However, the mutant did not exhibit decreased fluid accumulation in a rabbit ileal loop assay, suggesting that fluid accumulation in animals is not affected by lecithinase activity. Additionally, the phospholipase A (PhlA) in V.
mimicus was found to exhibit hemolytic activity against trout and tilapia erythrocytes and was cytotoxic to the fish cell line CHSE-214 [25]. Recently, the V. harveyi hemolysin (VHH) was shown to be a virulence factor during flounder infection and also had phospholipase activity on egg yolk agar [26]. Rock and Nelson [8] reported that the putative phospholipase gene (plp) from V. anguillarum exhibits 69% amino acid identity with the V. cholerae lec gene. Both plp and lec are located divergently adjacent to a hemolysin gene (vah1 and hlyA, respectively) [8,24]. Additionally, data strongly suggested that functional plp repressed transcription of its adjacent hemolysin gene, vah1, in V. anguillarum [8]. However, the enzymatic characteristics of Plp in V. anguillarum were not described.
Usually, phospholipases are divided into phospholipases A (A1 and A2), C, and D according to the cleavage position on target phospholipids. Most of lipolytic enzymes contain a putative lipid catalytic motif (GDSL) that was previously demonstrated in other bacterial and eukaryotic phospholipases [27]. However, Molgaard [16] proposed that four amino acid residues (SGNH) form a catalytic site, and are conserved in all members of the phospholipase family; therefore, phospholipases were re-named as a SGNH family. Multiple alignment analysis of 17 phospholipase homologues ( Fig. 1) demonstrates that V. anguillarum Plp belongs to SGNH hydrolase family, instead of GDSL family, since the GSDL motif was not fully conserved in these proteins (Fig. 1).
Recently, it was reported that mutation of the serine residue in the SGNH motif resulted in the complete loss of the phospholipase and hemolytic activities of VHH in V. harveyi [28] demonstrating the importance of this motif on the activity of phospholipase.
In contrast to the similarities of their catalytic motifs, the biochemical characteristics of bacterial phospholipases appear to be variable. For example, V. mimicus PhlA has a phospholipase A activity, which cleaves the fatty acid at either sn-1or sn-2 position, but no lysophospholipase activity [25]. Two phospholipases identified from mesophilic Aeromonas sp.serogroup O:34 show phospholipase A1 and C activity [29].
In addition, TLH of V. parahaemolyticus has PLA2 and lysophospholipase activity, and demonstrates a loss of activity at 55ºC for 10 min [20]. In this report, we show that V. anguillarum Plp has PLA2 activity, and is able to maintain activity at 64ºC for 1 h (Figs. 6 and 7). Therefore, the enzymatic characteristics of specific phospholipases are distinct even when they all belong to the SGNH hydrolase family (Fig. 1).
Phospholipases have been implicated in the pathogenic activities of a number of bacteria [30,31]. It is known that phospholipase activities often lead to cell destruction by degrading the phospholipids of cell membranes [30,32]. However, the relationships between phospholipases and virulence are not always clear. While the purified rPlp exhibits strong hemolytic activity against Atlantic salmon erythrocytes ( Fig. 7), Rock and Nelson [8] showed that a knock-out mutation of either the plp gene or the vah1 gene in V. anguillarum did not affect virulence of V. anguillarum during an infection study on juvenile Atlantic salmon. In this report, we show that when groups of rainbow trout are infected with either a plp mutant or a plp vah1 double mutant there is no significant difference in mortalities compared to fish infected with the wild type strain. Our data suggest that neither plp nor vah1 are major virulence factors during pathogenesis of salmonids. It was also reported that the deletion of lecithinase (Lec) activity in V. cholerae did not significantly diminish fluid accumulation in the rabbit ileal loop assay, indicating the lecithinase activity does not contribute significantly to enterotoxin activity [24]. In contrast, the direct IP injection of purified V. harveyi VHH protein caused the death of flounder with an LD 50 of about 18.4 µg protein/fish [26]. The rPhlA of V. mimicus also has a direct cytotoxic effect on the fish cell line CHSE-214 [25] suggesting that this phospholipase is a virulence factor during fish infection. In addition, the lecithinase purified from A. hydrophila (serogroup O:34) has been shown to be an important virulence factor to rainbow trout and mouse [29].
Generally, the hemolytic activity of phospholipases is dependent upon the hydrolysis of the phospholipids that reside in the erythrocyte membrane. Erythrocytes contain various phospholipids including phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), phosphatidylinositol (PI), and sphingomyelin (SM).
PC makes up 58% of the total erythrocyte phospholipids in the Atlantic salmon [33], but only 34% and 1% in rabbit and sheep erythrocytes, respectively [19]. Taken together with the high specificity of rPlp for PC ( Fig. 6), it was not surprising that rPlp was able to lyse the fish erythrocytes, but not sheep erythrocytes (Fig. 7), and that the plp mutant had decreased hemolytic activity on LB20-fish blood agar (Fig. 2). Our results are consistent were those reported for V. mimicus PhlA [25] and V. harveyi VHH [26], in which PhlA and VVH specifically lyse the fish erythrocytes.
We have previously reported that there are two hemolysin gene clusters in V.
anguillarum M93Sm, the vah1-plp cluster and rtxACHBDE cluster [9] and have described their regulation by H-NS and HlyU [17,34]. Mutation of both vah1 and rtxA results in the loss of all hemolytic activity on TSA-sheep blood agar [9], which is consistent with the data reported here that Plp has no activity on sheep erythrocytes.
We have also previously demonstrated that Plp is a putative repressor of Vah1, since mutation of plp increases vah1 expression by 2-3 fold [8]. In this report, we examined the hemolytic activity of various hemolysin mutants using freshly collected Rainbow trout blood ( Table 2) to investigate the relationships among three hemolysins of V. anguillarum. While culture supernatants from two of the three single mutants (JR1 and S123) and one of three double mutants (S183) exhibited ≥94%of the hemolytic activity as supernatants from the wild type strain M93Sm (

Methods
Bacterial strains, plasmids, and growth conditions. All bacterial strains and plasmids used in this report are listed in Table 1. V. anguillarum strains were routinely grown in Luria-Bertani broth plus 2% NaCl (LB20) [35] Allelic exchange mutagenesis. The allelic exchange rtxA mutation in V.
anguillarum S264 was made by using a modification of the procedure described by Milton et al. [25].The 5′region of rtxA was amplified using the primer pair pm256 and pm257 (Table3), digested with XhoI and XbaI, and then cloned into the region between the XhoI and XbaI sites on pDM4 (GenBank accession no. KC795686), deriving pDM4-rtxA5'. The 3′ region of rtxA was amplified using the primer pair pm258 and pm259 (Table3), digested with XbaI and SacI, and then cloned into the region between the XbaI and SacI sites on the pDM4-rtxA5'. The resulting pDM4-rtxA5'-rtxA3' was transformed into E. coli Sm10 to produce the transformant strain S252, which was mated with V. anguillarum S171 (vah1). Single-crossover transconjugants were selected with LB20 Kan 80 Sm 200 Cm 5 plates and, subsequently, double-crossover transconjugants were selected with LB20 Kan 80 Sm 200 5% sucrose plates. The resulting V. anguillarum colonies were transferred to TSA-sheep blood agar (Northeast Laboratories Service, Waterville, ME) and screened for non-hemolytic colonies (vah1 rtxA). The resulting colonies were checked for the desired allelic exchange using PCR amplification.
Complementation of mutants. The various mutants were complemented by cloning the appropriate target gene fragment into the shuttle vector pSUP202 (GenBank accession no. AY428809) as described previously by [8]. Briefly, primers (Table 3) were designed with EcoRI and AgeI sites and then used to amplify the entire target gene plus ~500 bp of the 5' and ~200 bp 3'flanking regions from genomic DNA of V. Hemolytic assays. The hemolytic activity of V. anguillarum strains was measured by two methods. First, single V. anguillarum colonies were transferred onto TSA-sheep blood agar, LB20-sheep blood agar (LB20 agar plus 5% sheep blood with heparin, obtained from Hemostat Laboratories) or LB20-fish blood agar (LB20 agar plus 5% rainbow trout or Atlantic salmon blood with heparin). Hemolytic activity of each colony was determined by measuring hemolytic zone surrounding the colonies after 24 h at 27°C. Additionally, the level of hemolytic activity was also quantitated using a microcentrifuge tube assay. The tubes contained 500 µl 5% erythrocytes (fish or sheep, suspended in 10 mM Tris-Cl, pH 7.5 -0.9% NaCl buffer) were mixed with 500 µl of bacterial supernatant or rPlp and incubated for 20 h at 27ºC. The samples were centrifuged at 1500 × g for 2 min at 4Cº, and the optical density of the resulting supernatant was read at 428 nm.

Phospholipase assay and thin-layer chromatography (TLC) analysis.
Phospholipase assays were performed in vitro with a BODIPY-phosphatidylcholine Fluorescence was detected and quantified using a Typhoon 9410 laser scanner.
Subcellular fractionation. V. anguillarum cells were fractionated as described previously [6] and the subcellular location of Plp determined. Briefly, 100 ml NSS-washed overnight grown bacteria cells were resuspended in 10 ml of ultrapure water for 20 min to cause osmotic shock and centrifuged (10,000 × g, 5ºC, 10 min) to collect the periplasmic fraction (the supernatant). The remaining pellets were washed twice with ultrapure water and lysed by sonication. Zealand White rabbits (Charles River Lab, MA). Briefly, 1 ml purified antigen (concentration = 100 µg/ml) was vigorously mixed with 1 ml TiterMax Gold adjuvant (Sigma) into a homogeneous suspension. About 10 ml of blood was withdrawn from the rabbits before immunization as a control. For the first injection, antigen-adjuvant mix was subcutaneously injected at 4 sites (over each shoulder and thigh; 100 µl/site).
The rabbits were boosted with single injections of antigen-adjuvant (100 µl) at day 28, 42, and 56. Blood was withdrawn 7-10 days after the 2 nd and 3 rd boosts to test the titer of antiserum using the western blot analysis. Antiserum with a high titer (> 1: 10,000) was aliquoted and stored at -70ºC.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and
Western blot analysis. Purified proteins or other protein samples were separated in 10% SDS-polyacrylamide gels. Prestained protein standards (Bio-Rad) and Laemmli sample buffer (Sigma) were used in all gels. Electrophoresis was performed at 100 V for 60-90 min. Gels were stained with either Coomassie blue G-250 or silver stain (Pierce, USA) to visualize the protein bands. Alternatively, proteins were transferred to nitrocellulose membranes for western blot analysis using the mini-  [29]. Briefly, V. anguillarum cells grown in LB20 supplemented with appropriate antibiotics for 22 h at 27°C were harvested by centrifugation (9,000 × g, 5 min, 4°C), washed twice in NSS, and resuspended in NSS (~2 × 10 9 cells ml -1 ).
Initial cell density was estimated by measurement of optical density at 600 nm. The actual cell density of NSS suspensions was determined by serial dilution and spot plating. All fish were examined prior to the start of each experiment to determine that they were free of disease or injury.    The zones of hemolysis were measured and the diameters were given in the figure.
This is a representative experiment from 3 replicate trials.

Abstract
The two hemolysin gene clusters previously identified in V. anguillarum, the vah1 cluster and the rtxACHBDE cluster, are responsible for the hemolytic and cytotoxic activities of V. anguillarum in fish. In this study, we used degenerate PCR to identify a positive hemolysin regulatory gene, hlyU, from the unsequenced V.

Introduction
Vibrio anguillarum is a marine member of the class Gammaproteobacteria. This highly motile gram-negative bacterium is the causative agent of warm-water vibriosis, a fatal hemorrhagic septicemic disease in fish, crustaceans, and bivalves (1). The mortality rate from V. anguillarum infections ranges from 30% to 100% (1).
Infections by these bacteria have resulted in severe economic losses to aquaculture worldwide (1,21) and affect many farm-raised fish including Pacific salmon, Atlantic salmon, sea bass, cod, and eel (1,4,5,21).
Hemolytic activity has been considered to be a virulence factor for V.
anguillarum and is thought to contribute to the hemorrhagic septicemia characteristic of vibriosis (7,16). Rock and Nelson (16) reported that the vah1 hemolysin gene cluster contains at least two genes, vah1 and plp, that affect hemolytic activity. Vah1 is a putative pore-forming hemolysin, which causes vacuolization of target cells (10).
It was suggested that pore-forming hemolysins, like HlyA in E. coli, cause direct lysis of blood cells by disrupting the membrane integrity (13). Mutations in the divergently transcribed plp result in both increased expression of vah1 and increased hemolysis, suggesting that Plp is a putative repressor of vah1 transcription (16).
Additionally, restoration of plp by complementation restores the wild-type level of vah1 transcription and hemolysis (16). Plp is a phosphatidylcholine (PC)-specific PLA2, which causes lysis of PC-rich fish erythrocytes (9).
Besides the vah1 cluster, a second hemolysin gene cluster, rtxACHBDE, was identified in the V. anguillarum (10). This gene cluster contains rtxA, which encodes a potent MARTX toxin and the specialized Type I Secretion System (T1SS) genes (rtxDBE) responsible for the secretion of the RtxA hemolysin/cytotoxin. A mutant containing mutations in both vah1 and rtxA completely lost hemolytic activity on sheep blood agar (10). Additionally, RtxA also exhibits cytotoxic activity and causes Atlantic salmon kidney (ASK) cells to round and die (10).
HlyU, a member of SmtB/ArsR family, is a metal-regulated transcriptional regulatory protein (17). It has been reported that HlyU is a positive regulator of   (20).
Degenerate primers (Table II) were designed from the conserved regions (Fig. 1), and used to amplify the possible hlyU gene from V. anguillarum M93Sm genomic DNA.
The PCR products were separated and purified from a 1% agarose gel and then subcloned into PCR2. into V. anguillarum M93Sm by conjugation (14). Transconjugants were selected by utilizing the chloramphenicol resistance gene located on the suicide plasmid. The incorporation of the suicide vector into the hlyU gene was confirmed by PCR analysis, as described previously (14). The resulting V. anguillarum hlyU mutant was designated S305 (Table 1)  RNA isolation. Exponential-phase cells (~0.5 × 10 8 CFU ml -1 ) and stationary-phase cells (2 × 10 9 CFU ml -1 ) of various V. anguillarum strains were harvested by centrifugation. Total RNA was isolated using the RNeasy kit (QIAGEN, USA) according to the manufacturer's instructions. All purified RNA samples were quantified spectrophotometrically by measuring absorption at 260 nm and 280 nm using a NanoDrop® spectrophotometer and stored at -75ºC for future use.
Semi-quantitative RT-PCR and Real-time RT-qPCR. Total RNA was isolated from exponential and stationary growth phase V. anguillarum cells as described above.
All RNA samples were treated with DNase and 100 µg RNA used as the template for reverse transcription (RT)-PCR. RT-PCR was performed using Brilliant SYBR Green single-step quantitative RT-PCR (qRT-PCR) Master Mix (Stratagene).
Briefly, gene-specific primers ( To identify the transcriptional start site, RNA was subjected to 5'-rapid amplification of cDNA ends (5'-RACE) using the 2 nd generation 5'-RACE kit (2). Primers used in RT-PCR are listed in Table 2. Briefly, 5 µg of RNA was used to generate specific first-strand cDNA from target mRNA (vah1, plp, rtxH, or rtxB) in a reverse transcriptase reaction with a gene specific primer. Poly(A) tails were added to the 3'-cDNA end using dATP and terminal deoxynucleotidyl transferase (or in some cases, poly(G) tails were added with dGTP and terminal deoxynucleotidyl transferase). A PCR product was amplified from the tailed cDNA by using a 5' RACE anchor primer (AP) ( Table 2) and the primer specific for that sequence. The PCR product was cloned into PCR2.1 cloning vector (Invitrogen, USA), and plasmids from appropriate transformants were purified and sequenced. DNase I protection assay. The DNA probes for the DNase I protection assay were amplified from V. anguillarum genomic DNA using PCR using primers (Integrated DNA Technologies, Inc.) shown in Table 3. Thus, two rtxH/rtxB intergenic region probes (4 and 5) were labeled with 6-FAM at 5' end on the upper strand and the lower strand, respectively. The two plp/vah1 intergenic probes (2 and 3) were also labeled with 6-FAM at 5' end on the upper strand and the lower strand, respectively. The assay was carried out using a method modified from Zianni et al (25

Identification of the hlyU gene in V. anguillarum. Previous studies
indicated that the hlyU gene is a conserved transcriptional regulator in many Vibrio species (11,12,22,23). We hypothesized that HlyU could be a putative regulator of the two hemolysin gene clusters in V. anguillarum. In order to identify the unknown hlyU gene in V. anguillarum, several hlyU genes from Vibrio species, including V.

Mutation in hlyU decreases hemolytic activity. An insertional mutation by
single crossover homologous recombination in the hlyU gene was obtained. The hemolytic activity of the hlyU mutant was determined and found to decrease about 5-fold compared with wild type strain M93Sm on sheep blood agar (Fig. 3).
Complementation of the hlyU mutant restored the hemolytic activity, which was even higher than wild type (Fig. 3), indicating that HlyU is a positive regulator of hemolysis in V. anguillarum.
Mutation in hlyU has no significant effect on cytotoxicity. One hemolysin gene, rtxA, has been shown to be a major virulence factor for V. anguillarum (10).
Previous studies revealed that RtxA has strong cytotoxic activity against Atlantic salmon kidney (ASK) cells, and causes cells to round-up, detach, and die (10).
However, experiments showed that ASK cells still rounded up and died when incubated with S305, M93Sm, or S307 cells (Fig. 4A) at a moi = 100 for 4 h, indicating the mutation in hlyU did not completely knock out the cytotoxicity of V.
anguillarum. Indeed, the LDH release assay revealed that S305 retained ~75-80% (P > 0.1) of cytotoxicity at all moi values compared to the wild type strain M93Sm (Fig.   4B) confirming that the mutation of hlyU had only a small, but statistically insignificant effect on cytotoxicity. As a negative control, the rtxA vah1 double mutant strain S183 exhibited no cytotoxicity compared to wild type strain M93Sm (Fig. 4B), confirming that rtxA and vah1 are the major cytotoxins in V. anguillarum (10). When strain S307 was assayed for cytotoxic activity by the LDH release assay, the activity was restored to same levels seen in M93Sm (Fig. 4B).
HlyU positively regulates hemolysin genes at the transcriptional level.  Table 4).
The data strongly suggest that HlyU is a positive regulator of rtx gene expression, playing an important role in the expression of rtx genes during both exponential and stationary phase. In fact, the data show that the mutation in hlyU has a larger effect on stationary phase expression of rtx genes than on exponential phase expression.
Additionally, expression of the same rtx genes increased to levels higher than wild type in the hlyU complement (Table 4) was 318 bases long (Fig. 6A). The +1 transcriptional start site (A) of plp is 73 bases prior to its start codon, with a predicted -35 and -10 promoter sequence of: TTGATT-N 13 -ATAAAT (Fig. 6A). The divergent hemolysin gene, vah1, had a transcriptional start site (G) 119 bases before the vah1 start codon, with a predicted -35 and -10 promoter sequence of TTGTGT-N 16 -TATTAA (Fig. 6A).
For the rtx gene cluster, the intergenic space between the divergent rtxH and rtxB genes is 325-bp. 5'RACE results show that the region between the transcriptional start sites of rtxH and rtxB is 187 bp (Fig. 6B). The +1 transcriptional start site (G) of rtxH is 103 bases prior to its start codon, with a predicted -35 and -10 promoter sequence of: TTGCGT-N 15 -TATAAT (Fig. 7B). The divergent rtxA transporter gene, rtxB, was found to have a transcriptional start site (C) 34 bases before the rtxB start codon, with a predicted -35 and -10 promoter sequence of TTGAGC-N 18 -TATAAT (Fig. 6B). Analysis of the predicted promoter regions of these two hemolysin clusters revealed strong similarities to a σ 70 consensus promoter: TTGACA-N 17 -TATAAT. Additionally, the putative ribosomal binding site (RBS) for all genes was also located upstream of the ATG start codons (Fig. 6A and B). intergenic region (Fig. 7B) and to fragment f of the plp-vah1 intergenic region (Fig.   7D). When unlabeled competitor DNA was added to each of these reactions, binding was decreased or abolished.
In an effort to more closely characterize the binding sites of HlyU for each hemolysin gene cluster, each of the two DNA sub-fragments that bound HlyU was examined by a DNase I protection assay as described in the Materials and Methods.
The results of these experiments revealed that HlyU protected an 18-bp region

Discussion
Hemolytic activity of V. anguillarum has been considered to be the virulence factor responsible for hemorrhagic septicemia during infection (1,5). We previously reported that there are two major hemolysin gene clusters in V. anguillarum M93Sm (10,16). The vah1 cluster consists of 4 genes, plp, vah1, llpA, and llpB gene. Vah1 is a putative pore-forming hemolysin, which shows strong homology to HlyA of V.
cholerae. HlyA integrates into the erythrocyte membrane to cause lysis (13). The genes encode proteins that are major hemolysins in the fish host (9). However, prior to this study, little was known about the regulation of these hemolysins in V. anguillarum.

It has been reported that HlyU regulates the expression of hemolysins in
Vibrio species. In V. cholerae, HlyU positively regulates the expression of hemolysin HlyA (23) and the HlyA co-regulated gene hcp (24). It was also suggested that mutation of hlyU attenuates the virulence of V. cholerae O17 in the infant mouse cholera infection model (22). Recent evidence suggests that HlyU is a master regulator of virulence in V. vulnificus, as several virulence factors, including vah1 and rtxA1, a homologue of rtxA of V. anguillarum, appear to be regulated by HlyU (8,11). Therefore, we hypothesized that the hlyU gene in V. anguillarum might encode a regulator for both hemolysin clusters in V. anguillarum. In this study, we used degenerate PCR to discover the unknown hlyU gene from the V. anguillarum genome. The experiment successfully identified an hlyU gene (Fig. 2) (22). In V. vulnificus, the LD 50 increased about 10 4 -fold in a hlyU mutant using the iron-overloaded mouse infection model (11) or the iron-normal mouse infection model (8). Additionally, cytotoxic activity was lost in an hlyU mutant of V. vulnificus (11). However, we found that in V. anguillarum cytotoxicity of the hlyU mutant remained relatively high according to both the LDH release assay and observations of morphological changes in ASK cells exposed to the hlyU mutant ( Fig. 4). These observations indicate that rtxA was still expressed in the hlyU mutant, even though rtxA expression was significantly decreased in the mutant ( Fig. 5 and Table 4). While our data indicate that HlyU is a positive regulator of rtxA, rtxH, and rtxB, these genes are still expressed in the absence of HlyU in V. anguillarum. It is interesting to note that transcription of rtxA, rtxH, and rtxB in the wild type strain and hlyU mutant all decrease during stationary phase (Table 4). This may suggest that either greater amounts of HlyU are required during stationary phase or that hlyU expression may be repressed during stationary phase.
Additionally, cytotoxicity data were consistent with the hemolytic activity assay, in which the hlyU mutant did not completely eliminate the hemolysis on the sheep blood agar (Fig. 3), indicating that the hemolysins were expressed in the mutant.
Interestingly, real-time RT-PCR data showed that the hlyU mutant did not effect the expression of vah1 and plp compared to the wild type strain (Table 4) vulnificus (12). It is reasonable to think that a similar situation might exist in the both hemolysin clusters of V. anguillarum.
In this study, transcriptional start sites of both hemolysin clusters were identified, and promoter regions for the potential HlyU binding were targeted (Fig. 6).
We found that the central regions of the intergenic sequence for each hemolysin gene cluster contains a conserved binding site for HlyU, as determined by both DNA mobility shift experiments (Fig. 7) and DNase I protection assays (Fig. 8). The two binding sites are quite similar (Fig. 8); the intergenic rtxH-rtxB protected binding region is 18 bp long, while the intergenic plp-vah1 region is 22 bp long and both have identical 5-bp direct repeats of TAAAA, strongly suggesting that HlyU binds as a dimer as suggested by Saha and Chakrabarti (17). In fact, the direct repeat may be a bit longer than 5 bp. If one uses an imperfect match, the direct repeat is 7-bp:

Manuscript III Introduction
Vibrio anguillarum is the causative agent of vibriosis, a fatal hemorrhagic septicemic disease. V. anguillarum infects more than 50 fresh and salt-water fish species including various species of economic importance to the larviculture and aquaculture industry, such as salmon, rainbow trout, turbot, sea bass, sea bream, cod, eel, and ayu (1). Infections by this bacterium have a mortality rate of 30% to 100% resulting in severe economic losses to aquaculture worldwide (2). anguillarum, while vah1 plays a more minor role in virulence.

The histone-like nucleoid structuring protein (H-NS) is a conserved global
regulator that belongs to a family of small nucleoid-associated proteins, including the factor for inversion stimulation (FIS), the heat-unstable protein (HU), and the integration host factor (IHF) (7). It is reported that function of H-NS is based on self-oligomerization and binding to DNA motifs to create DNA-protein-DNA bridges that can impede the movement of RNA polymerase (8). H-NS has been shown to repress expression of several virulence genes, including cholera toxin ctx (9, 10) and exopolysaccharide biosynthesis (vps) genes in V. cholerae (10,11), the RTX toxin gene (rtxA1) in V. vulnificus (12), and T3SS1 genes in V. parahaemolyticus (13 Transconjugants were selected by utilizing the chloramphenicol resistance gene located on the suicide plasmid. The incorporation of pNQ705-hns was confirmed by PCR amplification.

Construction of hns/hlyU double mutant. The hns/hlyU double mutant was
constructed by allelic exchange of hns, followed by insertional mutation of hlyU.
The allelic exchange mutation was made by using a modification of the procedure described by Milton et al. (18). Briefly, the plasmid pDM4 (GenBank accession no.
KC795686) was used to construct the hns::Km allelic exchange mutant as described previously (18). The 5′ region of hns was amplified using the primer pair pr40 and pr41 (Table 2), digested with XhoI and XbaI, and then cloned into the region between the XhoI and XbaI sites on pDM4. The 3′ region of hns was amplified using the primer pair pr42 and pr37 ( anguillarum mutants were checked for the desired allelic exchange using PCR amplification and then subjected to insertional mutation of hlyU as described above.
Complementation of the mutants. The various mutants were complemented by cloning the appropriate target gene fragment into the shuttle vector pSUP202 (GenBank accession no. AY428809) as described previously by Rock and Nelson (4).
Briefly, primers hns_comp(F) and hns_comp(R) ( All experiments were repeated at least twice.

Over-expression and purification of the V. anguillarum H-NS protein. The DNA
fragment encoding H-NS was PCR amplified by using Pm416 and Pm417 (Table 2) and cloned into a six-His tag expression plasmid, pQE30-UA (QIAGEN), generating the plasmid pQE-30 UA/H-NS (Table 1) DNase I protection assay. DNA probes for the intergenic region of each of the hemolysin gene clusters were amplified from V. anguillarum genomic DNA using PCR ( Table 2). Probes were labeled with 6-FAM at the 5' end of a certain strand of each probe (Fig. S1). The assay was carried out using a method modified from Li et al. (6). Briefly, 7.5 × 10 11

Mutation of hns increases hemolytic activity.
It was previously shown that V.
When the hemolytic activity of the hns mutant (M114, hns-) was tested on 5% TSA-sheep blood agar, it was found that mutation of hns resulted in increased hemolysis when compared to the wild type (M93Sm) (Fig. 2). Further, when the hns mutation was complemented (M116, hns+), hemolysis was reduced to levels below wild type (Fig. 2), suggesting that H-NS is a negative regulator of at least one of the two hemolysins, RtxA and Vah1. The lower hemolytic activity was probably due to the overexpression of H-NS since pSUP202 is a multicopy plasmid.  Table S1).

Mutation of hns increased cytotoxicity against
Further, complementation of the hns mutation down-regulated the expression of these genes back to or below wild type levels. The data strongly suggest that H-NS is a negative regulator of gene expression from both the rtxACHBDE and vah1-plp gene clusters. Hemolytic activity in the hlyU mutant decreased compared to that of wild type M93Sm, as previously reported by Li et al. (6). In contrast, hemolytic activity in the hlyU/hns double mutant increased over that in M93Sm and when hns was complemented in the double mutant, hemolysis decreased to levels seen in the hlyU mutant (compare Fig. 5A with Fig. 2). Changes in transcription of rtxA, rtxB, vah1, and plp corresponded with the changes in hemolysis (Fig. 5B and Table S1).

Mutation of hns
Specifically, transcription of each gene (rtxA, rtxB, vah1, and plp) increased in the absence of a functional hns and decreased in the presence of a functional hns.
The second set of determinations was carried out in cells lacking a functional hns: the hns mutant (hns-), the hns/hlyU double mutant (hns-/hlyU-), and the hns/hlyU double mutant with hlyU complemented (ES115, hns-/hlyU+); determinations of hemolytic activity and hemolysin gene expression were also done for wild type M93Sm (Fig. 6). In absence of a functional hns, hemolytic activity increased regardless of the presence or absence of hlyU ( Fig. 2 and Fig. 6A). Determination of hemolysin transcription by RT-qPCR corresponded with the hemolysis assay (Fig. 6B).
In absence of hns, rtxA and rtxB expression increased over levels in wild type cells.
For rtxA, all increases were >2-fold and significant (P<0.05). For rtxB, increases were small (generally <2-fold) and generally not significant. The presence or absence of hlyU had little or no effect (<2-fold) on rtxA and rtxB gene expression ( Fig.   6B and Table S1). Similarly, in the absence of a functional hns, expression of both vah1 and plp increased >9-fold in both exponential and stationary phase cells regardless of the presence or absence of a functional hlyU (Fig. 6B and Table S1).
As with the rtxACHBDE gene cluster, our data show that in absence of hns, the complement of hlyU (hns-/hlyU+) only exhibited minimal changes in expression of both vah1 and plp (around 2-fold) over the hns-strain and almost no change in vah1 and plp expression between the two strains ( Fig. 6B and Table S1). These data indicate that up-regulation of hemolysin genes by hlyU is hns-dependent. The results of these experiments revealed that rH-NS protected multiple regions in both rtxB/H and vah1/plp intergenic regions (Fig. 7). These regions are AT-rich (72-74% AT) and correspond to other H-NS binding sites described in other bacteria (21)(22)(23)(24)(25)(26)(27) (Fig. 8). The H-NS binding sites cover the promoter regions of all four genes (rtxB, rtxH, plp, vah1) with little or no overlap with the HlyU binding site in each of the intergenic regions (Fig. 8) (6). In the vah1/plp intergenic region, rH-NS bound to five sites, covering the -10 and -35 regions of both plp and vah1 promoters, but did not cover the HlyU binding site. In the rtxB/H intergenic region, rH-NS bound to six sites, covering the -35 region of both rtxB and rtxH promoters. In addition, rH-NS also bound to a seventh site just within the rtxB coding sequence. The rH-NS also was found to protect the three rtxB-proximal bases of the HlyU binding site (Fig. 8). the same as that found in strain M93Sm. Further, we saw no evidence of any tRNA genes, tranposases, interrupted genes, or pseudogenes in these surrounding regions.

H-NS binds
Finally, when we examined the codon usage patterns for the hemolysin genes and compared them to the chromosome in which each is found (rtx genes in choromosome I and plp and vah1 in chromosome II), no significant differences were found. These observations suggest that while the two hemolysin gene clusters are negatively regulated by H-NS, they were not horizontally acquired.
The hns mutant has attenuated virulence against rainbow trout. Since the expression of both hemolysin gene clusters is affected by H-NS, we tested the virulence of M93Sm, hns-, and hns+. Groups of ten rainbow trout were infected by IP-injection as described in the Materials and Methods with the wild type M93Sm.
hns-, or hns+ stains of V. anguillarum in NSS at a dose of ~4 × 10 5 CFU/fish, or with NSS only as a negative control. All M93Sm infected trout died by day 4, while 60% of hns-infected trout died by day 14. These results (Fig. 9) show that there was a significant difference (P = 0.005) in the virulence of the M93Sm wild type and the hns mutant. Complementation of hns restored virulence back to wild type levels with 90% mortality by day 4. Thus there was a significant difference (P = 0.029) in the virulence of the hns+ and hns-strains, and no significant difference (P = 0.413) between the wild type and hns+ strains.
At first glance, the decline in virulence for the hns mutant would appear to be counterintuitive, since both hemolysin gene clusters are up regulated in the hns mutant.
However, hns is considered important for bacterial fitness by properly regulating virulence and other genes during growth (26,27). To determine whether the loss of hns affected the growth of V. anguillarum, we tested the growth of M93Sm, hns-, and hns+ in LB20 (Fig. S3). While the three strains grew to nearly identical cell densities (OD 600 = 1.04 (M93Sm), 0.97 (hns-) and 0.97 (hns+) at stationary phase, hns-had a longer generation time compared to the wild type (58 min vs. 48 min, P <0.05). However, complementing the hns mutation did not result in a shorter generation time than the hns-(60 min vs. 58 min), suggesting there is no correlation between virulence in fish and fitness in LB20 for these three strains. 132

Discussion
Vibriosis caused by V. anguillarum has been recognized as a major problem for salmonid culture due to the significant economic loses it causes (29). While these bacteria use a variety of virulence factors including: iron transport/siderophore systems (30), the EmpA metalloprotease (31,32), motility (33, 34), lipopolysaccharides (LPS) (33,34), and exopolysaccharides (EPS) (35), it is the hemolysins/cytotoxins that directly kill host cells (4,5) and are thought to be the major contributors to the hemorrhagic septicemia that is characteristic of vibriosis (2).
Previously, we identified and described three hemolysin/cytotoxin genes in V.
anguillarum M93Sm, vah1 (4, 36), rtxA (5) and plp (4) (Li, Mou, and Nelson unpublished data). The three hemolysin genes (and associated transport genes) are organized into two gene clusters (Fig. 1). Additionally, both hemolytic activity and expression of the three hemolysin genes, vah1, plp and rtxA, are all higher in log phase than in stationary phase (6). Recently, we reported that HlyU positively regulates the expression of both hemolysin gene clusters by specifically binding to the vah1/plp and the rtxB/H intergenic regions (6).
In this study, we examined the role of H-NS in the regulation of hemolysin activity and gene expression in V. anguillarum M93Sm. Initially, the hns homologue in V. anguillarum was identified using the V. anguillarum M93Sm draft genome and an hns mutant and an hns complement strain were constructed. Mutation of hns resulted in increased hemolytic activity on 5% TSA-sheep blood agar, while complementation of the hns mutation reduced hemolysis to levels below wild type (Fig. 2). Mutation of hns also increased the cytotoxicity of both V. anguillarum culture supernatant (diluted 1:1 with PBS) and V. anguillarum cells (at MOI = 200) against ASK cells, while complementation of the hns mutation reduced cytotoxic activity (Fig. 3). Transcription of the three hemolysin genes (and related rtx genes) in the presence and absence of hns corresponded with hemolysin and cytotoxin activity, with increased transcription in the hns mutant and decreased transcription in the hns complement ( Fig. 4 and Table S1). These data show that H-NS is a negative  Table S1). These observations strongly suggest that in V. sites that extend towards the promoter sites (Fig. 8). The sites protected by rH-NS ( Fig. 8) are AT-rich. The five H-NS binding sites in the vah1/plp intergenic region have A+T% that range from 64.5% to 84.3 %, while the seven H-NS protected sites in the rtxB/H intergenic region have A+T% that range from 47.4% to 81.25%. In contrast, the flanking structural genes have much lower A+T%. The A+T% for plp = 57% and for vah1 = 55.7%; the A+T% for rtxACH = 51.6% and for rtxBDE = 54.5% ( Fig. 1). This reveals an interesting discontinuity between the structural genes and the intergenic regulatory regions. A similar discontinuity is also seen between the rtx structural genes and the intergenic regions in V. vulnificus and V. cholerae.
Additionally, that the structural genes have an A+T% nearly identical to the whole genome of V. anguillarum (55.49%) suggests that these virulence genes were not horizontally acquired. This is further supported by the observations detailed above that there is no evidence for any tRNA genes, tranposases, interrupted genes or pseudogenes in the 7.5 to 10 kbp of DNA flanking the hemolysin gene clusters.
Further, codon usage in the hemolysin genes is not significantly different from that of the chromosomes in which each gene cluster resides.             plp and vah1 and B) rtxH and rtxB were obtained by PCR amplification using the primers in Table 2.

B.
A. Figure S2. Supplement data for qRT-PCR analysis. Real-time qRT-PCR was performed to determine A) the expression level of hlyU in the wild-type strain (M93Sm), hns mutant (hns-), and the hns complement (hns+) during exponential and stationary growth phases. Each sample is the average of three replicates. The data presented are the averages of two independent experiments. B) Expression level of hns in the wild-type strain (M93Sm) during exponential and stationary growth phases.
Each sample is the average of three replicates. Statistically significant differences between samples are marked with a bracket and an * symbol. Error bars represent 1 standard deviation.   (5), which avoids the injection procedure frequently used in killed vaccines and make it possible to apply the vaccines to the larval or fry stages or some vulnerable species (3). Second, modified live vaccines are able to elicite humoral, mucosal and cell-mediated immunity (6), while killed or subunit vaccines are not effective at eliciting cellular responses (7). Third, modified live vaccines do not require an adjuvant, such as mineral oil, which is reported to cause side effects including decreased growth rates, chronic peritonitis, adhesions, granulomas and pigmentation in the peritoneal cavity (8). Fourth, due to omission of injection and adjuvant, modified live bacterial vaccines have a lower cost (7,9). Currently, there are eight licensed bacterial vaccines for fish in the United States. Five of them are killed vaccines and three of them are live cultures (Arthrobacter, Edwardsiella ictaluri and Flavobacterium columnare) (10). While there is a great need for modified live vaccines, modification of a specific pathogenic bacterium to result in a live vaccine strain may require a comprehensive understanding of its pathogenesis.
Similar results have also been observed for the uropathogenic Escherichia coli (UPEC), another intracellular bacterial pathogen (22). The TCA cycle-and gluconeogenesis-defective strains demonstrate significant fitness reductions during urinary tract infections (22). These observations suggest that the central metabolism of pathogens is important for pathogenesis.
The observations concerning the effects of the TCA cycle upon virulence allowed us to hypothesize that mutations in central metabolism-related genes, such as those that encode the TCA cycle enzymes could interrupt the infection process in fish.
To test the hypothesis, we identified and created six TCA cycle mutant strains plus one fructose metabolism mutant strain and tested their virulence against rainbow trout service (http://rast.nmpdr.org/rast.cgi) using the default settings (23).
Bacterial strains, plasmids and growth conditions. V. anguillarum strains (Table 1) were routinely grown in Lysogeny broth containing 2% NaCl (LB20) (24) (17). BLASTx was performed on the identified sequences against the NCBI non-redundant protein sequences database to confirm the accuracy of annotation.
anguillarum M93Sm were used to created insertional mutations in V. anguillarum M93Sm. Seven metabolism mutant strains were obtained and listed in Table 1. Similarly, fish infected with either sucA or sdhC mutants survived at 0% or 20%, respectively. In contrast, fish infected with the icd mutant had a significantly higher survival percentage (100%) compared to M93Sm (P = 0.013) (Fig 2B). The data indicate icd mutant is a highly attenuated or avirulent strain in these experimental conditions.
We then compared the virulence of M93Sm with the icd mutant by another infection route -immersion. Groups of 10 fish were infected by immersion as described in the Materials and Methods with either M93Sm or the icd mutant in a 1.5% sea salt solution at a dose of ~4 × 10 6 CFU ml -1 , or 1.5% sea salt solution without V.
anguillarum as a negative control (mock). Control experiments showed that neither the V. anguillarum cells nor the rainbow trout were adversely affected by the 1 h exposure to 1.5% sea salt solution ( Fig. S2A and 2C). During the 14-day experiment, 30% of M93Sm infected fish survived while 90% of icd mutant infected fish survived.
The difference is statistically significant (P = 0.007) (Fig 2C). Taken together, the two infection experiments demonstrate that the icd mutant exhibits highly attenuated virulence against rainbow trout.
Growth rates and growth yields of V. anguillarum wild type and mutant strains.
In order to determine the possible cause of attenuation in the icd mutant, we examined the growth curves of the wild type and the seven metabolism mutants of V.
anguillarum in one minimal medium (3M-Glucose) and two forms of rich media (LB20 and NSSM). The cell density in 3M-Glucose and LB20 was measured by spectrophotometry at 600 nm, while the cell density in NSSM was measured by viable count.
When cells were grown in Minimal Marine Medium plus 1% Glucose (3M-Glucose), the icd mutant and the sucA mutant failed to grow, since there is no alternative way to produce α-ketoglutarate or succinyl-CoA (Fig. 1). The sucC mutant, sdhC mutant, mdh mutant and cra mutant exhibited similar generation times to M93Sm (P = 0.139 to 0.985), while the fumA mutant exhibited a 33% longer generation time than that of M93Sm (P = 0.022) (Table S1).
In LB20 broth, M93Sm, icd mutant and cra mutant exhibited a classic bacterial growth curve with a lag phase, a log phase and a stationary phase, while the other strains sucA mutant, sucC mutant, sdhC mutant, fumA mutant and mdh mutant exhibited biphasic growth curves, with a lag phase then a log phase followed by slower growth (stationary/lag phase), then a second log phase, followed by a stationary phase (Fig 3A). The log phase in the first growth stage was named log phase I and the log phase in the second growth stage was termed log phase II. The generation times during the log phase(s) of all mutants (52.4 to 115.3 min) were longer than for M93Sm (44.0 min) (Fig 3B). virulence (22). The genes encoding the three hemolysins are expressed most strongly during log phase (15,16,30). To examine this possibility, we determined the expression of vah1, rtxA and plp during log phase using qRT-PCR (Fig 4A). Therefore, all metabolism mutants exhibit the same or higher levels of hemolysin gene expression compared to the wild type and attenuation of the icd mutant is not due to reduced expression from the hemolysin genes.
We have observed that Plp is the most efficient of the three hemolysins against fish erythrocytes (Li et al, unpublished data). Thus, zones of hemolysis on fish blood agar are primarily the result of Plp activity. The data indicate when compared to the wild type M93Sm, all TCA cycle mutants (but not the cra mutant) had significantly higher levels of hemolytic activity after 7 h of incubation on fish blood agar plates. After 23 h of incubation all TCA cycle mutants and the cra mutant had significantly larger zones of hemolysis on fish blood agar compared to M93Sm (P = 0.003 to 0.044) ( Fig   4B). The hemolysis activity results correspond with the expression data for plp and vah1 (Fig 4A).
Expression level of hlyU is higher in sucA mutant and mdh mutant than that in wild type. Previously, we demonstrated that the three hemolysin genes, vah1, rtxA and plp, are co-regulated by a repressor (H-NS) and an anti-repressor (HlyU) (15,30). In order to examine their possible roles in regulating hemolysin gene expression in the metabolism mutants, we examined the expression of hlyU and hns during log phase using qRT-PCR in two TCA cycle mutants -the sucA mutant and mdh mutant ( Fig   4C). Data show that both hlyU and hns were expressed at significantly higher levels in both the sucA and mdh mutant compared to M93Sm (P = 0.015 to 0.018). The expression of hlyU was 76% (in the sucA mutant) and 63% (in the mdh mutant) higher than in M93Sm, while the expression of hns was 14% (in the sucA mutant) and 36% (in the mdh mutant) higher than in M93Sm. Compared to the increase in hlyU expression, the increase of hns was relatively small (14% vs. 76% in sucA mutant, 36% vs. 63% in mdh mutant), suggesting the possibility that the relative increase in hlyU expression may be responsible for the increased expression of plp and vah1.
Pre-treatment by immersion with icd mutant protected rainbow trout from the subsequent challenge of V. anguillarum M93Sm. In order to test whether rainbow trout previously exposed to the highly attenuated icd mutant by immersion were protected against vibriosis by the fully virulent wild type M93Sm, we performed an infection challenge on fish that had been pre-treated with the icd mutant. A group of five fish that survived from the initial infection experiment by immersion (labeled as "pre-treated by immersion" in Fig 5A) and group of five "untreated" fish were immersion-infected by M93Sm at a dosage of ~4 × 10 6 CFU ml -1 and observed for 14 days. The pre-treated trout were challenged with the wild type cells 6 weeks after the initial infection with the icd mutant to allow adaptive immunity to develop. In the untreated group, no fish survived past day 2, while in the "pre-treated by immersion" group, all five fish survived for the entire 14-day observation period (Fig 5A). The difference between the two experimental groups is statistically significant (P = 0.003).
The results indicate that icd mutant elicits protective immunity when fish are inoculated by immersion.
Pre-treatment by injection of icd mutant elicited cross-serotype immunity against the subsequent challenge of V. anguillarum NB10. In order to test whether pre-treatment of fish with the icd mutant was able to provide cross-serotype protection, we carried out a second infection challenge experiment using the virulent wild type strain of V. anguillarum NB10 (serotype O1) with a serotype different from that of the M93Sm-derived icd mutant (serotype O2a), on fish that had been pre-treated with the icd mutant. A group of three fish that survived the initial infection experiment by immersion (labeled as "pre-treated by immersion" in Fig 5B), and group of five fish that survived from the initial infection experiment by injection (labeled as "pre-treated by injection" in Fig 5B) were infected by immersion with NB10 at a dosage of ~4 × 10 6 CFU ml -1 and observed for 14 days. As in the first challenge experiment, the pre-treated trout were challenged with the wild type cells 6 weeks after the initial infection with the icd mutant to allow adaptive immunity to develop. A group of four "untreated" fish were also included as a control. In the untreated group, no fish survived beyond day 4, while in the "pre-treated by immersion" group, one fish (33%) survived for the entire 14-day observation period. However, there is no statistically significant difference between the two experimental groups (P = 0.139). In the "pre-treated by injection" group, four fish (80%) survived for the entire 14-day observation period. When compared to the untreated control group, the number of survivors is statistically significant (P = 0.007) (Fig 5B). The results indicate icd mutant elicited cross-serotype immunity against V. anguillarum NB10 at least when applied by injection.

Discussion
An ideal modified live vaccine would be a strain that retains its invasive ability but does not cause further damage to the fish host, so that not only humoral but also cell-mediated immunity would be elicited, by simply applying the vaccine by immersion. For this purpose, factors that are exclusively required in the post-invasion stage, rather than the virulence factors that required for the invasive stage of the pathogen, would be the preferred targets to modify or knock out. We hypothesized that disrupting the central fueling pathways, such as the TCA cycle enzymes, were likely to result in defects that would accomplish this goal, especially since other investigators had success with this approach in both UTI E. coli (22) and Salmonella enterica (18)(19)(20). It has been suggested that intracellular pathogens inside host cell phagosomes are in nutritionally restricted sites and are likely to be starved for essential nutrients, especially when central fueling pathways are disrupted (20). However, unlike S. enterica and UTI E. coli, V. anguillarum is an extracellular pathogen and does not invade host cells. Thus, it was not clear that disrupting the central fueling pathways would necessarily result in V. anguillarum mutants unable to persist and grow in the fish host. To examine this question, we constructed 7 mutants -six TCA cycle mutants and one regulatory mutant (cra) and tested the mutants for virulence in rainbow trout and, subsequently, tested the attenuated mutant for its ability to protect the trout against challenge by virulent wild type strains of V. anguillarum.
For all metabolism mutants, there is no correlation between their virulence and growth rate in LB20, however, there is correlation between virulence and the maximum cell density since only the icd mutant had both the lowest virulence and the lowest maximum cell density (compare Fig 2A with Fig 3A), suggesting that a certain cell density is a prerequisite for a successful systemic infection. It is possible that once icd mutant cells get into the host fish, they are unable to grow to a high enough density to overwhelm the host innate immune defenses. Thus the host has time to mount an effective adaptive immune response. This raises the question as to whether icd mutant growth is restricted in the GI tract of the fish and/or after it penetrates the epithelial cell layer and invades the fish host. Our data suggest that the growth restrictions may be in both environments. Specifically, the icd mutant cells grew with nearly the same doubling time as the M93Sm wild type cells in NSSM, but their maximum cell density was only 24% that seen in M93Sm, suggesting that the cells may not reach a critical density necessary for successful invasion of the host.
However, we also injected fish with potentially lethal doses of the icd mutant, but all injected fish survived (unlike fish IP injected with M93Sm). This result suggests that even if icd mutant cells penetrate the epithelial layer and enter the fish host, they are unable to grow to a high enough density to overwhelm the host immune response.
All mutants constructed in this study exhibited the same or higher levels of hemolysin gene expression and higher hemolytic activity than M93Sm (Fig 4A and  4B), ruling out the possibility that the attenuated virulence of the icd mutant is due to the lack of hemolytic activity. Additionally, we suggest that the presence of secreted hemolysins will act to further stimulate the inflammatory response (31) and are likely to serve as antigens to elicit immunity against themselves, which is an advantage of this TCA cycle mutation approach over constructing avirulent strains containing mutations in specific virulence genes, such as the avirulent rtxA mutant obtained previously (14). The increase in expression of vah1-plp gene cluster (for at least the sucA and mdh mutants) is correlated to the increase in expression of hlyU, the transcriptional activator/antirepressor of the vah1-plp gene cluster (compare Fig 4A with Fig 4C). It is unclear how mutations in TCA cycle enzymes trigger increase expression of hlyU.
Immersion is the preferred method for vaccination in the aquaculture industry (3,5,32). In this study, rainbow trout pre-treated by immersion for 1 h with the icd mutant were fully protected from subsequent challenge with virulent wild type V. anguillarum M93Sm (Fig 5A). Moreover, since the TCA cycle exists in all aerobic organisms, as well as many facultative (like Vibrio) and anaerobic organisms, the TCA mutation approach is likely to be an easy method for the construction of modified live vaccines against a wide variety of pathogenic bacteria. Although there is a residual virulence of the icd mutant, a combination of icd and another TCA cycle gene mutation is expected to further attenuate the strain. In Samonella enterica, the combination of a sdhCDA deletion and frdABCD deletion resulted in further attenuation in virulence than either single gene deletion (19,20). 196 Cross-serotype protection is also highly preferred in vaccine development.
Data indicate the injection of icd mutant that derived from M93Sm (serotype O2a) had elicited cross-serotype immunity against the subsequent challenge of NB10 (serotype O1), demonstrating the potential of the TCA cycle mutation approach (Fig 5B).
Although the administration of icd mutant by immersion didn't show statistically significant protection against NB10 in this study, it should be noted that there were only three fish tested. More repeats need to be done to make a more solid conclusion.    (Table 1).   anguillarum strains grown in LB20 in log phase. *Statistically significantly higher than M93Sm.