The Effects of TCA Cycle Mutations on the Virulence of Vibrio Aguillarum Strains M93SM and NB10SM

Vibrio anguillarum is an extracellular bacterial pathogen that is a causative agent of vibriosis in finfish and crustaceans. Mortality rates range from 30% to 100% and systemic infection usually causes fish to die within 1-4 days of initial infection. The primary routes of infections are through the skin, gills and intestines. Chemotactic motility and the metalloprotease EmpA have been shown to be important virulence factors during the invasion stage while the siderophore anguibactin, flagellin subunits and lipopolysaccharides were shown to be important for persistence in the host during the post-invasion stage. Three secreted proteins that are cytotoxic against epithelial cells and erythrocytes have been characterized in V. anguillarum: the HlyA homolog Vah1, the phospholipase Plp, and the MARTX toxin RtxA. Previous research has demonstrated that mutations in vah1 and/or plp resulted in slight attenuation against juvenile Atlantic salmon (Salmo salar); however, rtxA mutants were avirulent. Expression of the cytotoxins are under control of the transcriptional activator HlyU and the repressor H-NS. Additionally, a V. anguillarum hns mutant showed attenuation in virulence when injected intraperitoneally, suggesting that proper coordination of gene expression is an important factor during the post-invasion stage. In manuscript I “Isocitrate dehydrogenase mutation in Vibrio anguillarum results in virulence attenuation and immunoprotection in rainbow trout (Oncorhynchus mykiss)”, seven central metabolism mutants were created in the M93Sm strain and characterized with regard to growth in minimal and complex media, expression of virulence genes and virulence in juvenile rainbow trout. Only the isocitrate dehydrogenase (icd) mutant was attenuated in virulence against rainbow trout challenged by either intraperitoneal injection or immersion. Further, the icd mutant was shown to be immunoprotective against wild type V. anguillarum experimental challenge. The icd mutant did not demonstrate a significant decrease in the expression of the three hemolysin genes was detected by qRT-PCR. Only the icd mutant exhibited a significantly decreased growth yield in complex media that was directly related to the amount of glutamate. A strain with a restored wild type icd gene was created and shown to restore growth to a wild type cell density in complex and minimal media and pathogenicity in rainbow trout. The data strongly suggest that a decreased growth yield, resulting from the inability to synthesize α-ketoglutarate derivatives (glutamate and glutamine), caused the attenuation despite normal levels of expression of virulence genes. Therefore, the ability of an extracellular pathogen to cause disease may be dependent upon the availability of host-supplied nutrients for growth. In manuscript II “Characterization of the growth and virulence of a Vibrio anguillarum citrate synthase mutant”, the role of glutamate auxotrophy during V. anguillarum M93Sm infection was further characterized. A citrate synthase (gltA) deletion mutant was created and characterized with regard to growth in minimal and complex media, expression of virulence genes, and virulence in juvenile rainbow trout. The ΔgltA mutant exhibited a decreased final cell density when grown in LB20 that resulted from the exhaustion of glutamate from the media. There was no significant decrease in the expression of the three hemolysin genes by the ΔgltA mutant when detected by qRT-PCR or mortality during challenge experiments. A ΔgltA mutant capable of growing in minimal media was isolated and shown to have a spontaneous mutation in the transcriptional activator of 2-methylcitrate synthase (prpR). This mutation resulted in an increase in expression of 2-methylcitrate synthase (prpC). The ΔgltA prpR(R66L) mutant was characterized with regard to growth in complex media and exhibited a growth advantage compared to the ΔgltA mutant after 24 h in spleen extract medium. Further, after growing 120 h in spleen extract medium, colonies of ΔgltA mutants were shown to be capable of growing in minimal media. ΔprpC and a ΔgltA ΔprpC mutants were created and characterized with regard to growth in minimal and complex media and virulence in juvenile rainbow trout. The ΔgltA ΔprpC mutant had no growth advantage in spleen extract medium compared to the ΔgltA but was still as virulent as the wild type against rainbow trout. As expected, the ΔprpC mutant was similar to the wild type in regards to both growth in minimal and complex media and virulence against rainbow trout. The data strongly suggests that simple starvation for αketoglutarate derivatives (glutamate and glutamine) is not directly linked to attenuation of virulence as previously proposed. Additionally, spontaneous mutations can occur that compensate for the original gene deletion if the new mutation can replace or bypass the lost metabolic reaction and results in a growth advantage. In manuscript III “Characterization of Vibrio anguillarum NB10Sm TCA cycle mutants” the role of central metabolism in virulence was examine in the O1 serotype strain of V. anguillarum NB10Sm. A V. anguillarum NB10Sm icd mutant was created, characterized for growth in complex media and demonstrated to be as virulent as the wild type in juvenile rainbow trout. Several additional central metabolism single and double mutants were created in the following genes cra, gltA, Δicd gltA, sucA, sucC, sdhC, ΔfrdA, ΔfrdA sdhC, and fumA and characterized with regard to growth in complex media. Two mutants (ΔsucA and ΔfrdA ΔsdhC) that demonstrated a significantly reduced growth yield compared to the wild type were further characterized with regard to their growth in several forms of complex media, expression of virulence genes, and virulence in juvenile rainbow trout. The data strongly suggest that there is no correlation between a lower growth yield in vitro and a decrease in virulence in vivo. Even though M93Sm and NB10Sm are same species, mutations made in the same TCA cycle genes can cause drastically difference results in regards to growth and virulence.

In manuscript I "Isocitrate dehydrogenase mutation in Vibrio anguillarum results in virulence attenuation and immunoprotection in rainbow trout (Oncorhynchus mykiss)", seven central metabolism mutants were created in the M93Sm strain and characterized with regard to growth in minimal and complex media, expression of virulence genes and virulence in juvenile rainbow trout. Only the isocitrate dehydrogenase (icd) mutant was attenuated in virulence against rainbow trout challenged by either intraperitoneal injection or immersion. Further, the icd mutant was shown to be immunoprotective against wild type V. anguillarum experimental challenge. The icd mutant did not demonstrate a significant decrease in the expression of the three hemolysin genes was detected by qRT-PCR. Only the icd mutant exhibited a significantly decreased growth yield in complex media that was directly related to the amount of glutamate. A strain with a restored wild type icd gene was created and shown to restore growth to a wild type cell density in complex and minimal media and pathogenicity in rainbow trout. The data strongly suggest that a decreased growth yield, resulting from the inability to synthesize α-ketoglutarate derivatives (glutamate and glutamine), caused the attenuation despite normal levels of expression of virulence genes. Therefore, the ability of an extracellular pathogen to cause disease may be dependent upon the availability of host-supplied nutrients for growth.
In manuscript II "Characterization of the growth and virulence of a Vibrio anguillarum citrate synthase mutant", the role of glutamate auxotrophy during V.
anguillarum M93Sm infection was further characterized. A citrate synthase (gltA) deletion mutant was created and characterized with regard to growth in minimal and complex media, expression of virulence genes, and virulence in juvenile rainbow trout.
The ΔgltA mutant exhibited a decreased final cell density when grown in LB20 that resulted from the exhaustion of glutamate from the media. There was no significant decrease in the expression of the three hemolysin genes by the ΔgltA mutant when detected by qRT-PCR or mortality during challenge experiments. A ΔgltA mutant capable of growing in minimal media was isolated and shown to have a spontaneous mutation in the transcriptional activator of 2-methylcitrate synthase (prpR). This mutation resulted in an increase in expression of 2-methylcitrate synthase (prpC). The ΔgltA prpR(R66L) mutant was characterized with regard to growth in complex media and exhibited a growth advantage compared to the ΔgltA mutant after 24 h in spleen extract medium. Further, after growing 120 h in spleen extract medium, colonies of ΔgltA mutants were shown to be capable of growing in minimal media. ΔprpC and a ΔgltA ΔprpC mutants were created and characterized with regard to growth in minimal and complex media and virulence in juvenile rainbow trout. The ΔgltA ΔprpC mutant had no growth advantage in spleen extract medium compared to the ΔgltA but was still as virulent as the wild type against rainbow trout. As expected, the ΔprpC mutant was similar to the wild type in regards to both growth in minimal and complex media and virulence against rainbow trout. The data strongly suggests that simple starvation for αketoglutarate derivatives (glutamate and glutamine) is not directly linked to attenuation of virulence as previously proposed. Additionally, spontaneous mutations can occur that compensate for the original gene deletion if the new mutation can replace or bypass the lost metabolic reaction and results in a growth advantage.
In manuscript III "Characterization of Vibrio anguillarum NB10Sm TCA cycle mutants" the role of central metabolism in virulence was examine in the O1 serotype strain of V. anguillarum NB10Sm. A V. anguillarum NB10Sm icd mutant was created, characterized for growth in complex media and demonstrated to be as virulent as the wild type in juvenile rainbow trout. Several additional central metabolism single and double mutants were created in the following genes cra, gltA, Δicd gltA, sucA, sucC, sdhC, ΔfrdA, ΔfrdA sdhC, and fumA and characterized with regard to growth in complex media.

Background
The aquaculture industry now produces half of all fish intended for human consumption and employs millions of people worldwide [1]. Although the first value sale of harvested fish has increased by 267% between 2004 and 2014 to over US$160 billion, infectious diseases, especially those caused by Vibrio spp. including Vibrio anguillarum, still represent a major impediment to the production of fish [1]. V. anguillarum causes diseases in crustaceans and bivalves, and is the leading causative agent of vibriosis in finfish including salmon, rainbow trout, turbot, sea bass, sea bream, cod, eel, and ayu [2].
Infections by this bacterial species have resulted in severe economic losses to aquaculture industries worldwide [3].
V. anguillarum is an extracellular pathogen that invades its host fish through the intestine, skin or gills [4,5]. Systemic infection by V. anguillarum usually causes fish to die within 1-4 days [6][7][8][9]. Chemotactic motility and the metalloprotease EmpA have been shown to be important virulence factors during the invasion stage while the siderophore anguibactin, flagellin subunits and lipopolysaccharides were shown to be important for persistence in the host during the post-invasion stage [2,10]. Three secreted proteins that are cytotoxic against epithelial cells and erythrocytes have been characterized in V.
Additionally, a V. anguillarum mutant that lacks H-NS, a global transcriptional regulator that represses the transcription of vah1, plp, and rtxA, showed attenuation in virulence when injected intraperitoneally, suggesting that proper coordination of gene expression is an important factor during the post-invasion stage [8].
Since the 1980s, several bacterial species that are auxotrophic for aromatic compounds have been shown to be avirulent [12][13][14][15][16]. More recently, mutants that are hypothesized to experience growth defects in the nutrient limited environment inside a phagocyte have been characterized. In Salmonella enterica, an intracellular bacterial pathogen, some tricarboxylic acid (TCA) cycle mutant strains were avirulent and immunoprotective for subsequent wild-type S. enterica infection [17][18][19][20][21] Accordingly, we hypothesized that mutations in central metabolism could interrupt the infection process of V. anguillarum in juvenile rainbow trout (Oncorhynchus mykiss). In this study, we identified and created six TCA cycle mutant strains plus one fructose metabolism mutant strain, and tested their virulence against juvenile rainbow trout using two infection methods, intraperitoneal (IP) injection and immersion. Further, the expression of each of the three hemolysin genes (vah1, plp, and rtxA) was examined to determine whether attenuation resulted from decreased virulence factor expression in these mutants. The growth rates and yield of each mutant strain in complex media were also determined. We specifically characterized the growth defect of the attenuated icd mutant. We also created, tested, and compared a restored wild type icd strain for virulence and growth to both the wild type and the icd mutant.

Methods
Bacterial strains, plasmids and growth conditions. V. anguillarum strains (Table 1) were routinely grown in Lysogeny broth containing 2% NaCl (LB20) [28]  Insertional mutagenesis. Insertional mutations were made by using a modification of the procedure described by Milton et al. [31]. Briefly, primers ( Briefly, V. anguillarum cells grown for 19 h at 27°C in LB20 supplemented with the appropriate antibiotics were harvested by centrifugation (9,000 × g, 5 min, 4°C), washed twice in NSS, and resuspended in NSS. Aliquots (100 µl) of the V. anguillarum NSS suspension were used to determine the OD600. The V. anguillarum NSS suspension was prepared to the desired specific cell density according to the conversion equation as determined by experimentation (data not shown): Cell density (10 8 CFU/ml) = 44.905 × OD600. The actual cell density of the suspension was confirmed by dilution and viable plate count. All fish were examined and determined to be disease and injury free prior to the start of each experiment. For IP injection, fish were anesthetized by tricaine methanesulfonate (Western Chemical, Ferndale, WA), (100 mg/l for induction and 52.5 mg/l for maintenance). V. anguillarum strains were IP injected into fish that were between 15 and 25 cm long in a 100 μl NSS vehicle at a dose of either 2 × 105 or 4 × 105 CFU/fish, or with NSS only as a negative control. For immersion, 10 ml of V.
anguillarum suspended in NSS, or 10 ml of NSS only as a negative control was added to a bucket filled with 10 L of water supplemented with 1.5% NaCl that was maintained at 18 ± 1ºC. Fish that were between 15 and 25 cm long were added and immersed for 1 h.
For both methods, fish inoculated with different bacterial strains were maintained in separate 10-gallon (38 L) tanks to prevent possible cross-contamination with constant water flow (200 ml/min) at 18 ± 1ºC. Death due to vibriosis was determined by the observation of gross clinical symptoms and confirmed by the recovery and isolation of V.
anguillarum cells resistant to the appropriate antibiotics from the spleen or head kidney of dead fish. Observations were made for 8-14 days. All fish used in this research project were obtained from the URI East Farm Aquaculture Center. All fish infection protocols were approved by the URI IACUC. (IACUC Protocol AN06-08-002).
RNA isolation. Exponential phase cells (~0.5 × 10 8 CFU/ml) of various V. anguillarum strains were treated with RNAprotect Bacteria Reagent (QIAGEN), following the manufacturer's instructions. Total RNA was isolated using the RNeasy kit and QIAcube (QIAGEN) following the instructions of the manufacturer. All purified RNA samples were quantified spectrophotometrically by measuring absorption at 260 nm and 280 nm using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific) and overall quality was assessed by gel electrophoresis. Samples were stored at -75ºC for future use. anguillarum NSS suspension was prepared to an OD600 of 0.420 (~4 × 10 7 CFU/ml) and diluted 1:100 into fresh media. Growth was monitored either by measurement of the OD600 or by serial dilution and plate counts.

Real-time quantitative RT-PCR (qRT-PCR
Resolving the merodiploid in the icd mutant. V. anguillarum icd mutant cells grown in LB20 supplemented with appropriate antibiotics for 19 h at 27°C were harvested by centrifugation (9,000 × g, 2 min), washed three times in NSS, and resuspended in NSS.
Cell suspensions (100 μl) were spread onto Marine Minimum Median (3M) + 0.15% glucose agar. Well-isolated colonies were picked and subsequently streak purified onto a new 3M + 0.15% glucose agar. Isolated colonies were then transferred to LB20-Cm 5 agar to screen for chloramphenicol sensitivity. Resolution of the merodiploid was confirmed by PCR amplification.

Statistical analysis.
A Kaplan-Meier survival analysis with log rank significance test was performed on the survival curves in the fish infection experiment. Student's T-tests assuming unequal variances were used for experiments containing two data groups. Oneway ANOVA with Tukey post hoc test was performed for all other experiments. P values of <0.05 were considered statistically significant.

Results
Identification and mutant construction of TCA cycle genes in V. anguillarum. In order to identify gene targets for mutagenesis the published genomes of V. anguillarum strains 775, 96F, M3, NB10, RV22, 90-11-286 and the V. anguillarum M93Sm draft genome (unpublished data) annotated by RAST were examined and found to have the following TCA cycle genes/operons: gltA, acnB, icd, sucAB, sucCD, sdhCDAB, frdABCD fumA, and mdh ( Fig. 1 and Table 3). While this set of genes allows for a fully functional TCA cycle, none of the strains have a fumC gene, which encodes the aerobic fumarate class II hydratase. All strains also lack the anaerobic fumarate hydratase (fumB) gene.
Additionally, all strains possessed cra, which encodes the repressor of fructose metabolism in S. enterica [22]. The V. anguillarum M93Sm sequences for the icd, sucA, sucC, sdhC, fumA, mdh, and cra genes were used to create insertional mutations in V.
anguillarum M93Sm. The seven mutant strains and the one restored strain listed in Table   1 were constructed using the primers listed in Table 2 as described in the Methods. icd mutant is highly attenuated for virulence against rainbow trout. The virulence of the seven V. anguillarum metabolism mutants were tested on rainbow trout and compared to wild type M93Sm in order to determine if mutations in metabolism could affect pathogenesis. Groups of five fish were infected by IP-injection (as described in the Methods) with either the wild type (M93Sm), icd mutant (XM420), sucA mutant (XM440), sucC mutant (XM450), sdhC mutant (XM460), fumA mutant (XM470), mdh mutant (XM410) or cra mutant (XM430) in NSS at a dosage of ~2×10 5 CFU per fish.
Injection with NSS only served as a negative control (Mock). During the 14-day observation window, 40% of M93Sm infected fish survived. Fish infected with the sucA mutant, sdhC mutant or icd mutant had a higher survival percentage than M93Sm (50% for sucA mutant, 80% for sdhC mutant, and 100% for icd mutant); however, only the difference between the icd mutant and M93Sm was statistically significant (p = 0.037) (Fig 2A). The experiment was repeated using a two-fold higher dose (~4×10 5 CFU per fish) of M93Sm and the three mutant strains (icd mutant, sucA mutant and sdhC mutant) that exhibited attenuated virulence in the previous experiment. At this dose, only 20% of M93Sm-infected fish survived. Only the icd mutant-infected fish had a statistically significant higher survival percentage (100%) compared to M93Sm (p = 0.0153) ( Fig   2B). The data indicate the icd mutant is avirulent in these experimental conditions. Further, we tested the virulence of M93Sm and the icd mutant by another infection route. Groups of 10 fish were infected by immersion as described in the Methods with M93Sm or icd mutant in 1.5% salt solution at a dose of ~4×10 6 CFU/ml, or just immersed in a 1.5% salt solution without V. anguillarum as a negative control (Mock). During the 14-day observation window, there was a statistically significant difference (p = 0.007) between the survival of M93Sm infected fish (30%) and icd mutant infected fish (90%) (Fig. 3). Taken together, the IP infection data and the immersion infection data demonstrate that the icd mutant is highly attenuated for infection in rainbow trout.
Pre-treatment by immersion with the icd mutant protected rainbow trout from the subsequent challenge of V. anguillarum M93Sm. Fish previously challenged by immersion with the icd mutant were subsequently challenged with the wild type M93Sm strain to test if the icd mutant was immunogenic. Six weeks after the initial infection, a group of five fish that survived the initial infection with the icd mutant (labeled as "treated with the icd mutant" in Fig. 4) and a group of five "untreated" fish were infected via immersion with M93Sm at a dose of ~4×10 6 CFU/ml and were observed for 14 days.
By day 2 all fish in the untreated group died. All fish in the group treated with the icd mutant survived the 14-day observation period. The difference between the two experimental groups was statistically significant (p = 0.008). The results indicate that the icd mutant is immunogenic and protective against wild type infection when administered by immersion.
All mutants exhibited either same or higher expression levels of the three hemolysin genes compared to wild type. Vah1, RtxA, and Plp are the three hemolysins found in M93Sm and are responsible for the hemolytic/cytolytic activity against fish erythrocytes, leukocyte and epithelial cells [11,7,9] and unpublished data]. We tested the expression of vah1, rtxA and plp during exponential phase to determine whether mutations in metabolism could affect the expression of these hemolysin genes. Data indicate that in all mutants except the icd mutant, expression of vah1 and plp were up regulated by 1. 49-16.15-fold compared to M93Sm with most of the changes being significant (Fig. 5). In the icd mutant, expression of plp was up regulated by 1.76-fold while the expression of vah1 was slightly decreased (to 49% of WT), neither of which was a significant change from M93Sm (Fig. 5). Plp is the most efficient hemolysin against fish erythrocytes [11].
TCA cycle mutants with an increased expression of plp also demonstrated an increased zone of hemolysis on 5% fish blood agar plates (Fig. S2). There was no change in the zone of hemolysis for the icd mutant. Expression of rtxA in all mutants was not significantly different from M93Sm (Fig. 5). Taken together, all metabolism mutants have the same or higher expression levels of hemolysin genes compared to the wild type.
icd mutant exhibited significant lower cell density limit than wild type in two forms of rich media. Fig. 6 shows the typical growth curves for the wild type V. anguillarum M93Sm and the seven metabolism mutants in LB20 broth. In these growth conditions, M93Sm, the icd mutant, and the cra mutant exhibited classic bacterial growth curves with a lag phase, an exponential phase and a stationary phase. The sucA, sucC, sdhC, fumA and mdh mutants all exhibited a two-stage growth curve, with each stage consisting of a lag phase and an exponential phase. The exponential phase in the first growth stage was named exponential phase I and the exponential phase in the second growth stage was named exponential phase II. The generation times of the exponential phases of all mutants were longer than for M93Sm ( Table 4). The final cell density (measured by OD600) of the icd mutant after 23 h was the lowest among all strains. Similarly, after 24 h of growth in LB20 the final cell density (CFU/ml) of the icd mutant was 47% that of M93Sm (Table 5) and the difference is significant (p = 0.011). M93Sm and the icd mutant were grown in NSS supplemented with 200 µg protein/ml of fish gastrointestinal mucus (NSSM) to better replicate conditions within a host. After 24 h of growth in NSSM the final cell density of the icd mutant was only ~31% of that for M93Sm (Table   5) and the difference is significant (p = 0.007).
Growth in LB20 supplemented with 118 mM glutamate restores growth of the icd mutant to wild type levels. The icd mutant is unable convert isocitrate into αketoglutarate, the immediate precursor of glutamate. Consequently, the icd mutant was only able to grow in 3M + 0.15% glucose with the addition of glutamate (Fig. 7A).
Glutamate was added to LB20 to determine if the icd mutant final cell density would increase. Fig. 7B shows the typical growth curves of M93Sm and the icd mutant in LB20 with (solid lines) and without (dashed lines) the addition of 118 mM of glutamate. After Resolving the merodiploid in the icd mutant restores growth and pathogenicity. A revertant to the wild type icd gene was selected to demonstrate that the icd mutant (XM420, a merodiploid with an insertion in the icd gene) contained no additional mutations that could be causing the loss of pathogenicity and decreased cell density.
Initially, attempts were made to complement the icd mutant in trans by cloning icd and its native promoter into the pSUP203 vector; however, all pSUP203-icd vectors isolated from E. coli SM10 contained single nucleotide substitutions that resulted in amino acid changes in icd that inactivated isocitrate dehydrogenase (data not shown). Since the icd mutant is unable to grow on 3M + glucose, icd mutants that spontaneously resolved the merodiploid were isolated on 3M + glucose agar plates as described in the Methods. The reversion rate of the icd mutant to a wild type phenotype grown in LB20 overnight was calculated to be 1 out of 1.6 × 10 10 cells. Fig. S1 shows the typical growth curves for M93Sm, the icd mutant and the restored icd strain in LB20 and 3M + 0.15% glucose.
M93Sm and the restored icd strain were able to grow in 3M + 0.15% glucose unlike the icd mutant ( Fig S1A). Additionally, when the strains were grown in LB20 the final cell density returned to wild type levels when icd was restored (Fig. S1B). To determine if restoring icd restores pathogenicity, juvenile rainbow trout were challenged via immersion with M93Sm, the icd mutant and the restored icd strain at a dose of between 4×10 6 and 8×10 6 CFU/ml. After day 8, 26% (5/19) of the M93Sm challenged fish, 40% (6/15) of the restored icd challenged fish and 95% (19/20) of the icd mutant challenged fish survived (Fig. 8). There was no statistically significant difference between M93Sm and the restored icd strain (P = 0.50). Again, there was a statistically significant difference between M93Sm and the icd mutant (p < 0.00004). The results indicate that when the merodiploid present in the icd mutant is resolved, wild type levels of growth in 3M + 0.15% glucose and LB20 and pathogenicity against juvenile rainbow trout is returned.

Discussion
The tricarboxylic acid (TCA) cycle is involved in the generation of energy through the oxidation of acetate. TCA intermediates serve as precursor metabolites for the synthesis of amino acids and peptidoglycan. The M93Sm genome along with the published genomes of V. anguillarum strains 775, 96F, M3,NB10,RV22, were examined for TCA cycle enzymes and the following genes were found: gltA, acnB, icd, sucAB, sucCD, sdhCDAB, frdABCD fumA, and mdh ( Fig. 1 and Table 3).
Additionally, cra, which encodes the repressor of fructose metabolism in S. enterica and E. coli and has previously been shown to be essential for S. enterica virulence, is present in the V. anguillarum genomes [22].
When in a nutrient limited environment, bacteria must be able to synthesize any essential metabolites that are not freely available in order to grow. Previous studies have shown that mutations in central metabolism genes result in attenuation of virulence in several intracellular pathogens including S. enterica, uropathogenic E. coli (UPEC), M.
tuberculosis and E. ictaluri [19, 21, 26, 23-25, 17, 18]. These observations suggest that central metabolism is necessary for these intracellular pathogens to function inside the nutrient-limited environment of the phagosome; however, V. anguillarum is not an  (Fig 2A and 2B) and 90% for immersion (Fig. 3). It is not thought that reversion of the merodiploid to a wild type phenotype caused the other metabolism mutants to be virulent because Cm resistant colonies were isolated from the organs of dead fish. IP injection bypasses the need for invasion. No mortalities resulted from IP injection with the icd mutant indicating that icd is required for V. anguillarum persistence and growth in fish tissues. Rainbow trout infected with the icd mutant via immersion and subsequently challenged with the M93Sm wild type showed 100% survival (Fig. 4) demonstrating that the icd mutant had immunoprotective effects and elicited an adaptive immune response. Moreover, as a proof of concept, the data suggest that an icd deletion mutant could be the basis for a live attenuated vaccine against V. anguillarum infection.
Our observation that a knockout of the icd gene results in attenuation of virulence raises the question of whether expression of required virulence genes is significantly reduced in the mutant and, therefore, results in attenuation. We previously identified and characterized three hemolysin/cytolysin genes and their encoded proteins secreted by V.
anguillarum: plp, vah1 and rtxA [11,7,9]. While mutations in plp and vah1 have modest effects on virulence against fish epithelial cells and fish, a knockout mutation in rtxA is avirulent in fish [11,7,9]. All metabolism mutants exhibited no significant declines in the expression of three hemolysins ( The growth rate and final cell density was determined for all metabolism mutants grown in LB20 for 24 h. The slowest growing mutant, fumA, was as virulent as the wild type while the mutant with the lowest final cell density, icd, was attenuated suggesting that decreased final cell density results in a loss of pathogenicity against rainbow trout (Fig. 2,  Table 4 and Table 5). When the mutation in icd was resolved, the restored icd strain demonstrated the wild type phenotype for both growth and pathogenicity ( The data also demonstrate the decreased growth yield was not do to a reduction of ATP production as addition of gluconate or succinate did not restore growth to a wild type cell density. It is interesting that the only other auxotrophic mutant, sucA, grows to a wild type cell density in LB20 and is as virulent as the wild type considering it cannot synthesize succinyl-CoA, a metabolite needed for the synthesis of lysine, methionine and diaminopimelic acid. Presumably, succinyl-CoA or its derivatives are not limiting in LB20 or in fish tissues. Furthermore, this also suggests that the icd mutant is primarily starved for glutamate and would not need to synthesize succinyl-CoA by metabolizing glutamate to α-ketoglutarate. We hypothesize that during infection the icd mutant is unable to obtain enough α-ketoglutarate derivatives to grow to a wild type cell density and, therefore, cannot reach a cell density necessary for a successful systemic infection.
In support, it has previously been demonstrated that a V. anguillarum M93Sm mugA mutant that was unable to grow in salmon intestinal mucus was avirulent against Atlantic salmon [39]. Additionally, when V. anguillarum 775 was cured of its plasmid-encoded siderophore, the mutant was unable to sequester iron and exhibited decreased virulence.
M93Sm is an O2α serotype and the presumed infection route is through the gastrointestinal tract as no necrotic skin lesions have ever been observed with this strain (unpublished data). The in vitro growth experiment (Table 5) suggests that there are not enough α-ketoglutarate derivatives in intestinal mucus to support the growth of the icd mutant to a wild type cell density even though it is the metabolite with the second highest concentration (3.03 mM) in rainbow trout mucus [42]. It should be noted that for in vitro growth experiments the concentrations of glutamate and glutamine in the mucus are not known and the growth conditions represent an ideal environment for growth; V. anguillarum does not have to evade the fish immune system or compete with commensal bacteria and it is not expected that the icd mutant will grow to the cell density shown in the in vitro growth experiments in the fish. As demonstrated by Muroga et al, [43] V.
anguillarum found in the spleen and intestine of moribund fish challenged via immersion only reached a cell density of 4.0×10 8 CFU/g and 2.5×10 7 CFU/g respectively. Altinok et al [44] showed a V. anguillarum succinate dehydrogenase mutant was avirulent against rainbow trout when injected at a dose at 10 5 CFU. Similar to our results, the authors showed that the succinate dehydrogenase mutant grew to a cell density slightly lower than the wild type at 12 h; however, the authors failed to show the growth yield at 24 h.
Further, the authors did not create a complement strain to demonstrate that the loss of virulence was solely do to mutating sdhB. Most importantly, the ATCC has redesignated their strain as a Pseudomonas species.

Conclusions
Seven V. anguillarum metabolism mutants were created and examined for pathogenicity against juvenile rainbow trout, hemolysin/cytolysin expression and growth in rich media. Of the central metabolism mutants, only the icd mutant showed strong attenuation in virulence, which did not result from a decrease in virulence factor expression. In addition, only the icd mutant had a final cell density that was lower than the wild type, which resulted from the inability to synthesize α-ketoglutarate and downstream metabolites. Taken together, the data suggest that during infection, if V.
anguillarum is unable to synthesize essential molecules (e.g. α-ketoglutarate/2oxoglutarate) and when those molecules or their derivatives (e.g. glutamate, glutamine) become limiting in the host, V. anguillarum will be unable to grow to a density necessary to sustain a systemic infection of the host.   Values calculated from data presented in Figure 6 during exponential growth. NA: not applicable   The arrows indicate the physiological directions of the reactions. The gene symbols of the enzyme for each reaction are listed beside the reaction. Boxed genes indicate the genes that were mutated in this study (Table 1).     auxotrophic for glutamate was demonstrated to be attenuated in virulence against juvenile rainbow trout. The inability of the mutant to synthesize essential metabolites (i.e. αketoglutarate and derivatives) was hypothesized to cause the attenuation. In this study, a citrate synthase mutant was created and characterized to determine if another mutant that is auxotrophic for glutamate would be attenuated.

Results:
A citrate synthase (gltA) deletion mutant was created and characterized with regard to growth in minimal and complex media, in vitro expression of virulence genes, and virulence in juvenile rainbow trout (Oncorhynchus mykiss). The ΔgltA mutant exhibited a decreased final cell density that resulted from the exhaustion of glutamate from the media. There was no significant decrease in the expression of the three hemolysin genes when detected by qRT-PCR or mortality during infection experiments.
A ΔgltA mutant capable of growing in minimal media was isolated and shown to have a spontaneous mutation in the transcriptional activator of 2-methylcitrate synthase (prpR).
This mutation resulted in an increase in expression of 2-methylcitrate synthase (prpC).
This ΔgltA prpR(R66L) mutant exhibited a growth advantage compared to the ΔgltA mutant after 24 h in spleen extract medium. Further, after growing 120 h in spleen extract medium, colonies of ΔgltA mutants were shown to be capable of growing in minimal media. ΔprpC and a ΔgltA ΔprpC mutants were created and characterized with regard to growth in minimal and complex media and virulence in juvenile rainbow trout. The ΔgltA ΔprpC mutant had no growth advantage in spleen extract medium compared to the ΔgltA mutant in spleen extract medium but unexpectedly, was still as virulent as the wild type against rainbow trout. The ΔprpC mutant was similar to the wild type in regards to both growth in minimal and complex media and virulence against rainbow trout.

Conclusions:
The data strongly suggests that simple starvation for glutamate will not directly result in attenuation of virulence. Additionally, spontaneous mutations can occur that compensate for the original gene deletion if the new mutation can replace or bypass the lost metabolic reaction and results in a growth advantage.

Background
Vibrio anguillarum is a causative agent of warm water vibriosis in finfish, crustaceans and bivalves. Morality rates from infection reach between 30% to 100% and result in severe economic losses to aquaculture industries worldwide [1,2]. Typically, systemic infection causes fish to die within 1 to 4 days [3][4][5][6].
V. anguillarum is an extracellular pathogen capable of invading its host through the gills, skin and intestines [7,8]. Numerous virulence factors that have been shown to be important during infection include extracellular proteases, hemolytic cytotoxins, iron acquisition systems (siderophores), lipopolysaccharides, chemotaxis, and flagella [9].
Mutations in vah1 and/or plp resulted in a slight attenuation in virulence against juvenile Atlantic salmon (Salmo salar); however, rtxA mutants cannot persist in host tissues and are avirulent against juvenile Atlantic salmon challenged via intraperitoneal injection [4,6,10].
Several bacterial species that are auxotrophic for aromatic compounds have been shown to be avirulent [12][13][14][15][16]. More recently, central metabolism mutants have been shown to be attenuated in pathogenicity in the intracellular pathogens Salmonella enterica, uropathogenic Escherichia coli (UPEC), Mycobacterium tuberculosis, and Edwardsiella ictaluri, [17][18][19][20][21][22][23][24][25][26][27]. The specific gene mutations that cause attenuation in virulence are hypothesized to reflect the metabolic reactions necessary to grow in the nutrient limited environment of the phagosome; if the pathogens cannot utilize the available nutrients they will fail to multiply to the threshold needed to cause disease [18,26]. Recently, a V. anguillarum M93Sm icd mutant was shown to be highly attenuated and immunoprotective in juvenile rainbow trout (Oncorhynchus mykiss) [28]. In vitro growth experiments demonstrated that the icd mutant grew to a lower cell density (~60%) than the wild type in two forms of rich media, due to limiting amounts of α-ketoglutarate derivatives (glutamate and glutamine) in the media. In this study, another TCA cycle mutant auxotrophic for glutamate was constructed by deleting the citrate synthase gene (gltA). The ΔgltA mutant was characterized for growth, hemolysin expression, and pathogenicity against juvenile rainbow trout. Additionally, a ΔgltA mutant strain containing a spontaneous mutation in the transcriptional regulator prpR capable of growing in 3M + 0.15% glucose (a medium which should be unable to support the growth of ΔgltA mutant) was isolated and characterized for growth in multiple forms of rich media composed of fish extracts that represent different stages of the infection process. Consequently, a ΔprpC (2-methylcitrate synthase) and a ΔgltA ΔprpC mutant were created and characterized for growth in multiple forms of media and pathogenicity against juvenile rainbow trout.

Methods
Bacterial strains, plasmids, and growth conditions. V. anguillarum strains (Table 1) were routinely grown in Lysogeny broth containing 2% NaCl (LB20) [29], LB20 + 118 mM glutamate, Marine Minimum Median (3M) + 0.15% glucose [30], NSSM (NSS supplemented with 200µg/ml of fish intestinal mucus) or Spleen extract medium supplemented with the appropriate antibiotic, in a shaking water bath at 27°C. Spleens, extracted from previously euthanized rainbow or brook trout, were added to 1 ml of NSS, placed in an ice-bath and sonicated using Fisher Scientific Sonic Dismembrator Model 500 four times at 10% power for a continuous 10 s. Tubes were then centrifuged (6000 × g, 10 min) and the supernatant was collected and centrifuged again (9000 × g, 10 min). 1 ml aliquots of the supernatant were added to a sterile 6-well plate (CytoOne) and sterilized with ultraviolet light using a Hoefer UV500 (Time = 10 min). An aliquot of the supernatant was added to LB20 and observed for growth to ensure that the spleen extract was sterile. E. coli strains (Table 1) were routinely grown in Lysogeny broth containing 1% NaCl (LB10) supplemented with the appropriate antibiotic, in a shaking water bath at 37°C. Antibiotics were used at the following concentrations: streptomycin, 200 g/ml (Sm 200 ); chloramphenicol, 20 g/ml (Cm 20 ) for E. coli and 5 g/ml (Cm 5 ) for V.

anguillarum.
Allelic exchange mutagenesis. Deletion mutations were made by using a modification of the procedure described by Milton et al. [31]. Genes of interested were identified using the M93Sm genome (Accession Number: NOWD00000000) previously annotated by RAST [33]. Briefly, primers ( plates, and subsequently, double-crossover transconjugants were selected with LB20 Sm 200 plates containing 5% sucrose. The resulting V. anguillarum mutants were checked for the desired allelic exchange by PCR amplification using primers (Table 2)  The V. anguillarum NSS suspension was prepared to an OD600 between 0.400 and 0.500 (~4 × 10 7 CFU/ml) and diluted 1:100 into fresh media. Growth was monitored either by measurement of the OD600 or by serial dilution and plate counts.
anguillarum strains were treated with RNAprotect Bacteria Reagent (QIAGEN) following the manufacturer's instructions. Total RNA was isolated using the RNeasy kit following the instructions of the manufacturer. All purified RNA samples were quantified spectrophotometrically by measuring absorption at 260 nm and 280 nm using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific) and overall quality was assessed by gel electrophoresis. Samples were stored at -75ºC for future use.

Results
ΔgltA mutant exhibited significant lower cell density limit than wild type in rich media and is auxotrophic for glutamate. Fig 1A shows a typical growth curve for the wild type (M93Sm) and the ΔgltA mutant composed of a lag phase, exponential growth phase and stationary phase grown in a rich medium (dashed lines). Like the previously described icd mutant, the gltA mutant has a lower cell density limit rich media [28]. The gltA mutant is auxotrophic for glutamate because it cannot run the first half of the oxidative TCA cycle which is needed to synthesize the immediate precursor molecule for glutamate production, α-ketoglutarate (Fig 1B). When glutamate (118 mM final concentration) was added to LB20 (solid lines), the ΔgltA mutant grew to a wild type cell density ( Fig 1A).
ΔgltA mutant exhibited either same or higher expression levels of the three hemolysin genes compared to wild type. As previously described, M93Sm secretes three hemolysins, Vah1, RtxA, and Plp that are most strongly expressed during exponential phase [11]. The expression of vah1, rtxA and plp was tested by qRT-PCR during exponential phase to determine if the ΔgltA mutant would have altered expression of these hemolysin genes. While there was no statistically significant increase in expression (P values ranged from 0.08 to 0.42), the gltA mutant exhibited 1.3-fold greater rtxA expression, 1.9-fold greater plp expression and 3.7-fold greater vah1 expression compared to the wild type (Fig. 2).
ΔgltA mutant was as virulent as the wild type against juvenile rainbow trout. Ten juvenile rainbow trout were challenged via immersion at doses of 3×10 6 to 4×10 6 CFU/ml with the wild type (M93Sm) and the ΔgltA mutant. There was no significant difference in survival between M93Sm (10%) and the ΔgltA mutant (20%) (P = 0.31) (Table 3); however, it took three days for the ΔgltA mutant to reach 80% mortality compared to two days for 90% mortality in fish infected with M93Sm. While the data strongly suggest that deleting gltA had a very small effect on virulence, the delay in mortalities raised the possibility that a mutation that compensated for the loss of citrate synthase was selected in the infected fish.
A mutation in prpR allowed the ΔgltA mutant to grow in 3M + 0.15% glucose. In order to test for the possibility that a compensatory mutation that would bypass the ΔgltA mutation could have been selected, the ΔgltA mutant was grown in 3M + 0.15% glucose for multiple days. An Δicd mutant was also created and was unable to grow in 3M+0.15% glucose (data not shown). The ΔgltA mutant began to grow after 48 h, reaching wild type levels by 72 h (Fig. 3). A frozen stock was created from the 72 h culture. A single colony, isolated from the frozen stock, was used for another growth experiment in 3M + 0.15% glucose. This new strain was able grow in 3M + 0.15% glucose during a 24 h incubation. When the gltA gene was sequenced from this strain, there was no change in the sequence from the original ΔgltA mutation; it remained truncated.
It has previously been described that 2-methylcitrate synthase (PrpC), an enzyme that combines propionyl-COA and oxaloacetate to synthesize 2-methylcitrate, can act in a promiscuous manner and synthesize citrate from acetyl-CoA and oxaloacetate [34].
Accordingly, prpB (2-methylisocitrate lyase), prpC, prpR (transcriptional regulator) and the intergenic regions were sequenced using the primers listed in Table 2 (Table 4). In LB20 and NSSM, both the ΔgltA mutant and the ΔgltA prpR(R66L) mutants grew to a total cell density that was statistically lower than M93Sm. The difference between the ΔgltA mutant and the ΔgltA prpR(R66L) mutant was not statistically significant. In the spleen extract medium, M93Sm grew to the highest cell density (7.7×10 7 CFU/ml) followed by the ΔgltA prpR(R66L) mutant (1.5×10 7 CFU/ml), while the ΔgltA mutant grew to 5.7×10 7 CFU/ml. The difference between M93Sm and the ΔgltA prpR(R66L) mutant was not statistically significant. The data strongly suggest that the ΔgltA prpR(R66L) mutant had a growth advantage in spleen extract medium compared to the ΔgltA mutant.
The prpR(R66L) mutation in the ΔgltA mutant can be selected for in spleen extract medium. The wild type (M93Sm), the ΔgltA mutant and the ΔgltA prpR(R66L) mutant were grown in spleen extract medium for multiple days (Fig. 5). By 96 h, number of viable cells in the cultures of both M93Sm and the ΔgltA prpR(R66L) mutant began to decline. However, the cell density of the ΔgltA mutant increased between 96 h and 120 h.
Colonies were chosen at random from the ΔgltA mutant 120 h agar plate and screened to ensure that the ΔgltA mutation was still deleted and if the colonies could grow in 3M + 0.15% glucose. All colonies had the truncated version of gltA while 4 out of the 6 colonies grew in 3M + 0.15% glucose with only a 24 h incubation. Although, it was not screened, the ability to grow 24 h in minimal media (Fig. 5) was presumably from a mutation in prpR. The data suggest that the prpR(R66L) mutation in the ΔgltA mutant strain is favored in spleen extract medium.
ΔgltA ΔprpC mutant is unable to grow in 3M + 0.15% glucose and grows to a lower cell density limit in two forms of rich media. A ΔprpC mutant and a ΔgltA ΔprpC mutant were created and their growth was examined. After 24 h, the ΔgltA ΔprpC mutant grew to a lower cell density in LB20 compared to the wild type (M93Sm) (Fig. 6A) and failed to grow in 3M + 0.15% glucose after 192 h (Fig. 6B). Growth in spleen extract medium was monitored and after 210 h the ΔgltA ΔprpC mutant failed to grow to the cell density of M93Sm or the ΔgltA mutant (Fig. 6C). The ΔprpC mutant grew to a wild type cell density in all forms of media (Fig. 6). The data demonstrate that when prpC is deleted in the ΔgltA mutant, the bacteria are no longer able to grow in 3M+0.15% glucose and lose their growth advantage in spleen extract media. However, the loss of prpC alone had little effect on growth.
ΔgltA ΔprpC double mutant is as virulent as the wild type against juvenile rainbow trout. Five to nine juvenile rainbow trout were challenged via immersion at a dose of 4×10 6 CFU/ml to 6×10 6 CFU/ml with either the wild type (M93Sm), the ΔprpC mutant, or the ΔgltA ΔprpC mutant. Fish infected with either M93Sm or the prpC mutant had 0% survival after 3 and 2 days, respectively (Table 5). While fish infected with the ΔgltA ΔprpC mutant had 22% survival, there was no significant difference in survival between M93Sm (0%) and the ΔgltA ΔprpC mutant (P = 0.23). The data suggest that deleting both gltA and prpC causes only a small attenuation of virulence. Aromatic compound auxotroph mutants have been demonstrated to be highly attenuated since the 1980s [12][13][14][15][16]. More recent studies have demonstrated that intracellular pathogens containing TCA cycle mutations are attenuated for virulence if the specific mutations prevent growth using the available nutrients in the nutrient-poor phagosome [17][18][19][20][21][22][23][24][25][26][27]. Since V. anguillarum is not an intracellular pathogen and cannot survive in macrophages, it would not be subject to this type of nutrient limitation [36]. Additionally, the initial infection site V. anguillarum M93Sm, an O2α serotype, is presumably the intestines (external necrotic lesions have never been observed in fish exposed to this strain, unpublished observations) where it grows in the glutamate rich intestinal mucus [7,8,37]. Therefore, the previously described link between glutamate auxotrophy and attenuation in virulence in M93Sm warranted further investigation [28].

Discussion
A ΔgltA strain was created and showed the same growth phenotype as the previously described icd mutant ( Fig. 1) [28]. Unlike the icd mutant, the gltA mutant was as virulent as the wild type against juvenile rainbow trout. However, death from systemic infection was delayed by about 1 day (Table 3) suggesting that a compensatory mutation that could bypass the citrate synthase deletion was being selected for in the ΔgltA mutant.
Indeed, a ΔgltA strain capable of growth in 3M + 0.15% glucose media (a medium in which it should fail to grow) was isolated (Fig. 3). When this ΔgltA strain was sequenced a spontaneous mutation was found in prpR, the transcriptional activator of 2methylcitrate synthase (prpC), that changed arginine 66 to a leucine (this strain was named ΔgltA prpR(R66L)). The normal enzymatic function of PrpC is to synthesize 2methylcitrate from propionyl-CoA and oxaloacetate; however, it has been demonstrated that PrpC can act promiscuously, substituting acetyl-CoA for propionyl-CoA to synthesize citrate [34]. The ability of the ΔgltA prpR(R66L)) mutant to grow in 3M + 0.15% glucose media resulted from an increase in expression of prpC (Fig. 4). The increase in expression of prpC enabled the cells to bypass the gltA deletion, an observation previously described in an E. coli ΔgltA strain by Digianantonio et al [35].
The hypothesized route of infection for V. anguillarum M93Sm is through the anus of the fish where it damages the lining of gastrointestinal tract to enter the circulatory system where it accumulates in the spleen [7,8]. Accordingly, the in vitro growth of the ΔgltA prpR(R66L) mutant was characterized in NSSM and spleen extract medium to represent the environment of early and mid-stage infections respectively (LB20 served as a control). The growth advantage of the ΔgltA prpR ( The data presented here suggest that our previously proposed hypothesis that simple starvation for glutamate results in attenuation of virulence must be modified. While the small decrease in mortality during infection of the ΔgltA ΔprpC mutant suggests glutamate auxotrophy may contribute to attenuation, it may not be the single cause of attenuation in virulence of the icd mutant. The oxidative branch of the TCA cycle is at the intersection of three regulatory molecules: acetyl-CoA, citrate and αketoglutarate. Citrate accumulation has been shown to be detrimental to the growth rate and growth yield of E. coli with spontaneous citrate synthase mutants outgrowing icd or aconitase (acnA and acnB) double mutants during in vitro growth experiments in complex media [38,39]. Additionally, in Staphylococcus aureus, citrate has been shown allosterically activate CcpE, a regulator of metabolism and virulence factors [40]. An E.
coli icd mutant was shown to have increased expression of the glyoxylate shunt causing lower acetyl-CoA production [41]. During acetogenesis, phosphotransacetylase converts acetyl-CoA into acetyl-P, a global regulator that is used to phosphorylate two-component signal transduction pathways [42]. Acetyl-P has been shown to induce virulence factor expression in V. cholerae, S. enterica and E. coli [43][44][45][46]. Further, acetyl-CoA accumulation alone, without being converting to acetyl-P, was shown to cause increased expression of the V. cholerae virulence gene activator ToxT [47]. Lastly, α-ketoglutarate has been shown to be a master regulator that coordinates nitrogen and carbon metabolism by inhibiting the production of cyclic AMP and has been shown to regulate virulence factors in UPEC [48][49][50] anguillarum.

Conclusion
A V. anguillarum ΔgltA strain was created and was demonstrated to be similar to the previously described icd mutant; both strains demonstrated a decreased final cell density in multiple forms of rich media and had expression of hemolysin genes that was similar to the wild type. Unlike the icd mutant, the ΔgltA mutant was as virulent as the wild type in juvenile rainbow trout. Spontaneous mutations arose in prpR that caused an increase in expression of the promiscuous enzyme prpC allowing the ΔgltA mutant to bypass the gltA deletion and grow in minimal media and have a growth advantage in spleen extract media. A ΔgltA ΔprpC mutant was created and was demonstrated to be unable to grow in minimal media and no longer have a growth advantage in spleen extract media. However, the ΔgltA ΔprpC mutant exhibited only a slight attenuation in virulence compared to the wild type against rainbow trout further suggesting that simple starvation for a required nutrient (i.e. glutamate) will not directly result in attenuation of virulence as previously proposed.  Values in brackets represent growth percentage compared to wild type.
** statistically significant to wild type (P < 0.01), * statistically significant to wild type (P < 0.05)    picked from the 120 h plate and used to inoculate 3M + 0.15% glucose. In addition, colony PCR using primers flanking gltA was performed to ensure gltA was still deleted.

Background
Vibrio anguillarum is a causative agent of warm water vibriosis, which causes a fatal hemorrhagic septicemic disease of fish, as well as morbidity and mortality in crustaceans and bivalves [1][2][3]. Mortality rates from vibriosis range between 30 to 100% and contribute to significant economic losses to the aquaculture industry worldwide [1,3]. Typically, the initial site of infection is the skin, gills or intestines and in vivo studies have demonstrated that death from system infection usually occurs 1-4 days post infection [4][5][6][7][8][9]. Numerous virulence factors that have been shown to be important during infection include extracellular proteases, hemolytic cytotoxins, iron acquisition systems (siderophores), lipopolysaccharides, chemotaxis, and flagella [10].
Previously, a V. anguillarum M93Sm icd mutant was shown to be highly attenuated against juvenile rainbow trout and immunoprotective [11]. Auxotrophic or TCA cycle mutants attenuated in virulence have been created in several intracellular bacteria species [12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]. The attenuation of the V. anguillarum M93Sm mutant was hypothesized to have resulted from the mutant being unable to grow to a final wild type cell density. The decreased growth yield resulted from the cells being auxotrophic and starved for glutamate [11]. If this same growth limitation were to occur during an infection, the cells would not be able to surpass threshold needed to cause disease.
Interestingly, unlike the previously reported attenuated auxotrophic or TCA cycle mutants, V. anguillarum is not an intracellular pathogen and would not be experience the sdhC, fumA and cra were created by using previously constructed suicide vectors [11].
Additional insertional mutations were made by using a modification of the procedure described by Milton et al. [36]. Briefly, primers ( Allelic exchange mutagenesis. Deletion mutations were made by using a modification of the procedure described by Milton et al. [36]. Suicide vectors were constructed via restriction enzyme digestion/ligation (frdA, sucA, icd, CUO) or Gibson Assembly (sdhC).

Cell growth experiments. V. anguillarum cells grown overnight at 27°C in LB20
supplemented with the appropriate antibiotics were harvested by centrifugation (9,000 × g, 2 min), washed twice and resuspended in in NSS. A 200 µl aliquot of the V.
anguillarum NSS suspension was transferred into a 96-well plate with a clear flat bottom and the optical density at 600 nm (OD600) was read by a VersaMax™ Absorbance Microplate Reader (Molecular Devices). The V. anguillarum NSS suspension was prepared to an OD600 between 0.300 and 0.350 (~1.5 × 10 9 CFU/ml) and diluted 1:100 into fresh media. Growth was monitored either by measurement of the OD600 or by serial dilution and plate counts. Total RNA was isolated using the RNeasy kit following the instructions of the manufacturer. All purified RNA samples were quantified spectrophotometrically by measuring absorption at 260 nm and 280 nm using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific) and overall quality was assessed by gel electrophoresis. Samples were stored at -75ºC for future use.

Results
icd mutant grew to the same cell density limit as wild type in complex media. A V.
anguillarum M93Sm icd mutant was previously shown to grow to a lower cell density than the wild type in nutrient rich media (e.g. LB20 and NSSM) [11]. A V. anguillarum NB10Sm icd insertional mutant was constructed and Fig. 2A shows a typical growth curve for the wild type (NB10Sm) and the icd mutant. Unlike the previously described M93Sm icd mutant, the NB10Sm icd had a slower growth rate during mid and late exponential phase growth compared to the wild type but reached the same cell density after growth for 24 h ( Table 3, Supplementary Fig. 4) [11]. No additional copies of icd are found in the genome and the NB10Sm icd mutant was unable to grow in minimal media indicating that there is no promiscuous enzyme that can act as an isocitrate dehydrogenase (Fig. 2B). To ensure the icd insertional mutant (a merodiploid) did not resolve to restore the wild type, an icd deletion strain was created and showed no difference in growth compared to the icd insertional mutant (Fig. 2C). The data suggest that unlike the M93Sm icd mutant, the NB10Sm icd mutant can obtain enough αketoglutarate derivatives (e.g. glutamate, glutamine, and peptides containing those amino acids) from LB20 to grow to a wild type final cell density.
icd mutant exhibits at a wild type growth rate when LB20 is supplemented with gluconate or glutamate. Fig. 3 shows the typical growth curve of the wild type (NB10Sm) (black) and the icd mutant (blue) in LB20 not supplemented (Fig. 3, dashed lines) or supplemented with 101.7 mM gluconate (Fig. 3A, solid lines) or 118 mM glutamate (Fig. 3B, solid lines). The data indicate that the icd mutant exhibited a growth rate during the mid and late exponential growth phases similar to the wild type when LB20 was supplemented with a metabolite that feeds into central metabolism upstream (gluconate) or downstream (glutamate) of the icd mutation (Table 3). Interestingly, the wild type grew to a higher cell density than the icd mutant when LB20 was supplemented with gluconate. The data suggests that under these conditions the icd mutant will utilize gluconate to grow rapidly but will experience a cell density limitation when the cells exhaust the available glutamate.  Supplementary Fig. 2). The CUO region was deleted to determine if it enabled the NB10Sm icd mutant grow to a final cell density that was similar to the wild type. Fig. 4 shows a typical growth curve of the wild type (NB10Sm), the icd mutant, the ΔCUO mutant and the ΔCUO icd mutant. The ΔCUO mutant had a growth curve nearly identical to the wild type and the ΔCUO icd mutant had a growth curve that was nearly identical to the icd mutant indicating that the presence of the CUOs did not enable the NB10Sm icd mutant to reach to a wild type cell density after 24 h.
icd mutant is as virulent as the wild type against juvenile rainbow trout. To determine if the NB10Sm icd mutant would be as virulent as the wild type (NB10Sm), 10 juvenile rainbow trout were challenged via immersion at a dose of between 6.5×10 6 CFU/ml to 9.4×10 6 CFU/ml (Table 4). Although it took seven days for the icd mutant to cause 70% morality and five days for the wild type to cause 80% mortality, there was no statistically significant difference in virulence between the two strains (P= 0.30).
sucA mutant, fumA mutant, and a ΔfrdA sdhC mutant grow to a final cell density in complex media lower than the wild type. It was previously proposed that a decreased final cell density in complex media resulting from a nutrient limitation in vitro could correlate to attenuation in virulence in vivo [11]. Since the NB10Sm icd mutant did not show either a decrease in final cell density or a loss of virulence, an additional ten central metabolism single and double mutants were created in the following genes: cra, gltA, Δicd gltA, sucA, sucC, sdhC, ΔfrdA, ΔfrdA sdhC, and fumA (Table 1). Fig. 5A shows a typical growth curve of the wild type (NB10Sm) and various central metabolism mutants.
Examination of the growth yield for the various mutant strains after 24 h by optical density shows that the sucA mutant (OD600 = 0.551), fumA mutant (OD600 = 0.618), and a ΔfrdA sdhC mutant (OD600 = 0.482) failed to grow to a wild type (OD600 = 1.08) cell density (Fig. 5B). Although the ΔfrdA ΔsdhC mutant cannot convert succinate to fumarate, it is not auxotrophic for any metabolite. Fig. 6C shows the 24 h cell density (OD600) of the wild type and ΔsucA mutant grown in LB20 or LB20 supplemented with lysine (118 mM) or DL-α, ɛ-diaminopimelic acid (118 mM) or lysine, DL-α, ɛ-diaminopimelic acid, and methionine (118 mM). The ΔsucA mutant is auxotrophic for the TCA cycle intermediate succinyl-CoA, a metabolite that is needed in diaminopimelic acid, lysine and methionine biosynthesis [40]. The addition of lysine, DL-α, ɛ-diaminopimelic acid and methionine to LB20 did not restore the growth of the ΔsucA mutant to a wild type cell density and indicates that the cells are either unable to import lysine or DL-α, ɛ-diaminopimelic acid or methionine or the cells require metabolites upstream of diaminopimelic acid in the biosynthetic pathway of lysine and the enzymes involved in this pathway cannot run in reverse. Interestingly, the addition of lysine and DAP actually caused the wild type to grow to a lower cell density, indicating that these cells do not favor these metabolites for growth.  (Table 6). In all forms of complex media, the ΔsucA mutant grew to the lowest cell density followed by the ΔfrdA ΔsdhC mutant.

The
All mutants exhibited either same or lower expression levels of the three hemolysin genes compared to wild type. Vah1, RtxA, and Plp are three hemolysins secreted by V.
anguillarum M93Sm that have been characterized and determined to be responsible for the hemolytic/cytolytic activity against fish erythrocytes, leukocyte and epithelial cells [7,9,41] and unpublished data]. It is unknown how these hemolysins contribute to the virulence of V. anguillarum NB10Sm. The expression of vah1, rtxA and plp during exponential phase was tested to determine whether mutations in metabolism could affect the expression of these hemolysin genes (Fig. 7A). rtxA was the most highly expressed (>4.6×10 3 copies/10 ng RNA) of the three hemolysins in NB10Sm and there was no statistically significant difference between the wild type (NB10Sm) and the two mutants (ΔsucA and ΔfrdA ΔsdhC). The mutants have a decrease in expression in both plp and vah1 compared to the wild type. However, these genes were expressed at the limit of detection by qRT-PCR. In the wild type, there were <30 copies/10 ng RNA for plp and <10 copies/10 ng RNA for vah1.
All mutants exhibited lower levels of empA metalloprotease gene expression compared to wild type. It has previously been demonstrated that V. anguillarum NB10 empA metalloprotease mutants are slightly attenuated in virulence and that empA expression is dependent upon RpoS and, therefore, highly expressed during stationary phase [6,42,43]. The expression of empA during stationary phase was tested to determine whether mutations in metabolism could affect the expression of this gene ( would be as virulent as the wild type (NB10Sm), 5 juvenile rainbow trout were challenged via immersion at a dose of between 1.0×10 6 CFU/ml to 4.0×10 6 CFU/ml (Table 7). There was no statistically significant difference in mortality between fish challenged with the wild type (100%) and the ΔsucA mutant (100%) (P= 0.39) or ΔfrdA ΔsdhC mutant (100%) (P= 0.65). The data once again indicate that central metabolism mutants that have a lowered final cell density due to a starvation of a nutrient will not necessarily be attenuated in virulence.
Regardless of the cause of the attenuation of the M93Sm icd mutant, it was still immunoprotective and has potential as a live cell vaccine against vibriosis. V.
Accordingly, this study originally sought to create a live attenuated O1 serotype strain by mutating icd in V. anguillarum NB10Sm. The NB10Sm icd mutant (the insertional mutant and the deletion mutant) did not show the cell density limitation previously observed in the M93Sm icd mutant in complex media (Fig. 2) nor was it reduced in virulence against juvenile rainbow trout (Table 4) [11]. The NB10Sm icd mutant (ES920 and ES926) did have a longer generation time than the wild type during mid and late exponential growth that was shortened by the addition of gluconate or glutamate to LB20.
The data suggest that the decrease in growth rate of the NB10Sm icd mutant (ES920) anguillarum NB10 compared to V. anguillarum M93Sm in noninduced (LB20) conditions [6]. LB20 is composed of yeast extract and tryptone (both of which are composed of peptides of varying sizes), although bacteria can only utilize peptides that are six amino acids in size or smaller [44,45]. The increase in expression of empA could allow the NB10Sm icd mutant (ES920) to more efficiently break down and utilize proteins and large sized peptides than M93Sm, preventing the cells from being starved for glutamate.
Since the NB10Sm icd mutant (ES920) was not attenuated in virulence and did not have a final cell density limit that was lower than the wild type after 24 h of growth in complex media, additional central metabolism mutants and double mutants were constructed (Table 1). Mutations in fumA, sucA or ΔfrdA sdhC caused reductions in the final cell density (Fig. 5). Focus was placed on the sucA mutant and the ΔfrdA sdhC mutant and deletion mutants were constructed for each strain to prevent the mutants from spontaneously resolving the merodiploid (a phenomenon that was previously exploited to create the icd revertant strain [11]). The ΔfrdA ΔsdhC mutant can run the reductive and oxidative branch of the TCA cycle to synthesize any intermediate and accordingly, is not auxotrophic for any metabolite when it is grown with glucose as the only carbon source.
However, the data suggest that the cells favor metabolites that feed into the TCA cycle at α-ketoglutarate such as glutamate and glutamine. Since the TCA cycle can only run αketoglutarate to succinate (the reaction catalyzed by icd is nonreversible), the cells are starved for the amino acid aspartate. The growth limitation of the ΔsucA mutant was not fully characterized but is presumed to have resulted from the inability of the cells to synthesize succinyl-CoA, a metabolite utilized in the biosynthetic pathways for the production of diaminopimelic acid, lysine and methionine [40]. The initial hypothesized route of infection for V. anguillarum NB10Sm is through the skin, gills or anus of the fish where it enters the circulatory system and accumulates in the spleen [4,5]. Accordingly, the final cell densities of the ΔsucA mutant and ΔfrdA ΔsdhC mutant were determined in media that represent early infection (NSSM and NSSSkM) and mid-infection (spleen extract medium). In all forms of complex media (including LB20 as a control) the ΔsucA mutant grew to the lowest cell density, followed by the ΔfrdA ΔsdhC mutant ( Table 6).
The data suggests that, even under ideal conditions at each point of the infection process, the mutants could be starved for essential metabolites in vivo. However, the infection experiments (Table 7) demonstrate that the starvation does not correlate to a loss of virulence in vivo (discussed below).
The expression of the virulence factors (rtxA, plp, vah1, empA) were examined in the ΔsucA mutant and ΔfrdA ΔsdhC mutant and compared to the wild type (Fig. 7). NB10 has been previously been shown to be non-hemolytic on blood agar, presumably from the hemolysins being degraded by EmpA before they can lyse the red blood cells [9]. This study expands upon this explanation and shows that the actual expression levels of the hemolysins is much lower in NB10Sm compared to M93Sm. Expression of rtxA during log phase was relatively similar between the wild type and the mutants, although a decrease in expression was observed for plp and vah1 (Fig. 7). This decrease was not believed to be relevant since the expression of plp and vah1 was very low in the wild type (28 and 6.5 copies per 10 ng of total RNA, respectively). It has previously been shown that a V. anguillarum M93Sm rtxA mutant is avirulent [7]. It is not known how rtxA contributes to the virulence of V. anguillarum NB10Sm especially since the RtxA of M93Sm and NB10Sm have different effector domains [46]. This study demonstrates that the expression of rtxA is 127-fold lower than in M93Sm [Spinard,EJ et al,in preparation,chapter 2 of this manuscript]. Yet, the expression of NB10Sm rtxA was still 160-fold higher than the expression of NB10Sm plp suggesting that it could be important for the infection process of NB10Sm. Expression of the metalloprotease empA was examined during stationary phase. empA has been shown to be nonessential for the virulence of V.
anguillarum NB10 [42,47]. It has previously been demonstrated that the expression of empA is cell density dependent [6]. Accordingly, the ΔsucA mutant had the lowest final cell density (13% of the wild type) and the lowest expression level of empA (0.03% of the wild type). The transcript of VanT, the transcriptional activator of empA, is stabilized by the sigma factor RpoS [43,47]. Alternatively, the decrease in expression of empA in the mutants could have resulted from a decrease in expression of RpoS.
Regardless of the lowered cell density limit and the decrease in empA production, both the ΔsucA mutant and ΔfrdA ΔsdhC mutant were as virulent as the wild type in juvenile rainbow trout ( Table 7). The metabolite starvation that prevents the mutants from reaching a wild type growth yield in vitro does not prevent the cells from reaching the threshold needed to cause disease in vivo. Accordingly, the approach that was used in this study to find attenuated strains, screening central metabolism mutants that grow to a lower cell density than the wild type in vitro, was unsuccessful probably because V.

Oxaloacetate Citrate
Isocitrate  Statistical analysis was based on data at 24 h. Error bars represent 1 standard deviation.  and NB10Sm grown in LB20: ** P < 0.01, *** P < 0.001 and **** P < 0.0001. Error bars represent 1 standard deviation. Error bars represent 1 standard deviation.   Vibrio coralliilyticus RE22 is a causative agent of vibriosis in larval bivalves. We report the draft genome of V. coralliilyticus RE22 and describe additional virulence factors, which may provide insight into the mechanism of pathogenicity.

Succinyl-CoA
Vibrio coralliilyticus RE22 (formally Vibrio tubiashii RE22) is a marine pathogen and a causative agent of vibriosis in larval bivalves (1). Disease is characterized by high mortality rates leading to severe loss of production in shellfish hatcheries (2)(3)(4).
The genome of V. coralliilyticus RE22 encodes two extracellular metalloproteases besides the previously described vtpA and vtpB. One protease shows similarity to epp in Vibrio anguillarum (12) while the other contains a domain conserved in M4 family of metalloproteases (13)(14)(15)(16)(17). In addition to vthA, three putative hemolysin/cytolysin genes were discovered. A putative MARTX toxin operon encoding three T1SS transport proteins, a MARTX toxin, and a hypothetical protein is on the megaplasmid. Unlike typical MARTX toxin gene clusters, the transporter genes are not transcribed divergently from the MARTX toxin (18). Instead they seem to be in the MARTX operon, upstream of the MARTX toxin gene. Unlike most MARTX toxin gene clusters, no rtxC (acyltransferase) is present in the operon. Additional putative hemolysins include a phospholipase/hemolysin located on chromosome 2 that shows similarity to plp in V.
This Whole Genome Shotgun project has been deposited in DDBJ/ENA/GenBank under the accession no. LGLS00000000. The version described in this paper is the first version, LGLS01000000.

Funding information. This work was supported by an award from the Rhode Island
Science and Technology Council to DRN and DR. This research is based in part upon work conducted using the Rhode Island Genomics and Sequencing Center, which is supported in part by the National Science Foundation under EPSCoR Grants Nos.
0554548 & EPS-1004057. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Vibrio tubiashii subsp. europaeus is a bivalve pathogen isolated during episodes of mortality affecting larval cultures in different shellfish hatcheries. Here we announce the draft genome of the type strain PP-638 and describe potential virulence factors, which may provide insight into the mechanism of pathogenicity.
Vibrio tubiashii subsp. europaeus is an emerging bivalve pathogen identified recently as the etiological agent responsible of larval and spat mortalities in clam, oyster and abalone cultures detected in Spanish and French hatcheries (1,2). This pathogen is a causative agent of vibriosis inducing mass mortalities and important economic losses, representing the main bottleneck for the production process in shellfish aquaculture (1,2). Genomic DNA was sequenced using an Illumina MiSeq at the Rhode Island Genomics and Sequencing Center at the University of Rhode Island. Reads were trimmed using the CLC Genomics Workbench (v8.5.1) for quality, ambiguous nucleotides and adapters.
2,943,708 paired-end and 3,234,516 mate-paired reads providing 199× coverage were assembled using Spades (v3.1.1) using the default parameters (1). Contigs were filter based on 34× coverage and 4000 bp length resulting in ten contigs with an N50 of 1,788,614 and an average G + C content of 45.37%. The assembly was mapped to Vibrio tubiashii ATCC 19109 using the CLC Microbial Genome Finishing module and resulted in six contigs mapping to chromosome 1, one complete contig representing chromosome 2, one complete contig representing the p251-like megaplasmid and one contig mapping to the p57-like plasmid (2). One 4,885 bp contig did not map to the reference genome.
The draft genome was submitted to Rapid Annotations using Subsystems Technology (RAST) for annotation resulting in 5157 open reading frames (3)(4)(5).
Encoded on chromosome 2 of the V. tubiashii subsp. europaeus PP-638 T genome is a putative metalloprotease that has a 75% similarity to VtpA found in Vibrio coralliilyticus RE22 (6). Another protease that has a 71% similarity to Epp in Vibrio anguillarum M93Sm is encoded on chromosome 2 (7). There are three putative hemolysins and phospholipases encoded in the genome. One hemolysin located on chromosome 2 has a 67% similarity to Plp in V. anguillarum M93Sm (8,9). In V.
anguillarum M93Sm, plp is divergently transcribed from the pore-forming hemolysin/cytolysin vah1 (9). In V. tubiashii subsp. europaeus PP-638 T , the Plp homolog is also divergently transcribed away from a pore-forming cytolysin, though it has a 42% similarity to aerolysin in Aeromonas eucrenophila, not vah1 (NCBI Reference Sequence: WP_042642875.1). The genome encodes two secretion systems (T3SS and T6SS) that are used to deliver effecter molecules directly into the host. The T3SS secreted virulence factor has a domain that is similar to the GTPase-activating domain found on YopE from Yersinia pestis (10)(11)(12)(13)(14). While the T6SS structural components are encoded on the p251like megaplasmid, the protein responsible for forming the puncturing tip of the T6SS secretion system, VgrG, appears to be encoded by two genes. One VgrG-encoding gene is on Chromosome 1 and the second is on Chromosome 2.
Nucleotide sequence accession numbers. This Whole Genome Shotgun project has been deposited in DDBJ/EMBL/GenBank under the accession number LUAX00000000.
The version described in this paper is the first version LUAX01000000.