Construction and Evaluation of Recombinant Attenuated Edwardsiella piscicida Vaccine (RAEV) Vector System Encoding Ichthyophthirius multifiliis (Ich) Antigen IAG52B

We have successfully designed and constructed a RAEV vector system with regulated-delayed attenuation in vivo attributes that synthesizes Ichthyophthirius multifiliis (Ich) protective antigen IAG52B to enable vaccination of fish susceptible to edwardsiellosis and white spot disease. The first feature of this vaccine delivery system is an Edwardsiella piscicida strain carrying genomic deletions of asdA. AsdA is an enzyme necessary for the synthesis of diaminopimelic acid (DAP), which is an essential component of the peptidoglycan layer of the cell wall of Gram-negative bacteria. asdA mutant strains have obligate growth requirements for DAP in the medium or a plasmid vector with the wild-type asdA gene enabling synthesis of DAP. This balanced-lethal plasmid vector-host system in E. piscicida enables as a second feature the synthesis of recombinant antigens to induce protective immunity against fish pathogens. Recombinant protective antigen IAG52B from the fish pathogen I. multifiliis was synthesized by RAEV strains harboring the AsdA+ plasmid pG8R8029. The third feature of this vaccine strain is a regulated-delayed attenuation in vivo phenotype that is based on the replacement of an arabinose-regulated araC ParaBAD cassette for the promoters of the fur and crp genes of E. piscicida such that the expression of these genes is dependent on arabinose provided during growth. Thus, following colonization, the Fur and Crp proteins stop being synthesized due to the lack of arabinose and attenuation is progressively achieved in vivo to prevent generation of diseases symptoms. Our vaccine strain χ16022 with the genotype ΔasdA10 ΔPfur170::TT araC ParaBAD fur ΔPcrp68::TT araC ParaBAD crp contains the AsdA+ plasmid, pG8R8029, which encodes the IAG52B antigen. Vaccine strain χ16022(pG8R8029) is attenuated and induces systemic and mucosal IgM titer against E. piscicida and Ich in zebrafish. In addition, transcript levels of tnf-α, il-1β, il-6 and il-8 were significantly increased in different tissues of vaccinated zebrafish compared to unimmunized fish. Zebrafish vaccinated with χ16022(pG8R8029) showed 60% survival upon intracoelomic (i.c.) challenge with a lethal dose of virulent E. piscicida strain J118. Our RAEV system could be used as a generalized vaccine-vector system to protect teleost fish against multiple bacterial, viral and parasitic infectious diseases.

different tissues of vaccinated zebrafish compared to unimmunized fish. Zebrafish vaccinated with c16022(pG8R8029) showed 60% survival upon intracoelomic (i.c.) challenge with a lethal dose of virulent E. piscicida strain J118. Our RAEV system could be used as a generalized vaccine-vector system to protect teleost fish against multiple bacterial, viral and parasitic infectious diseases.

INTRODUCTION
Edwardsiella piscicida is a Gram-negative, facultative anaerobic bacterium of the family Enterobacteriaceae. It is the causative agent of edwardsiellos in many economically important species of fish including catfish (Ictalurus furtatus), tilapia (Tilapia nilotica), European eel (Anguilla anguilla) and Indian major carp (Catla catla) (1)(2)(3). Vaccination would be an effective method to prevent and control Edwardsiella outbreaks. Liverecombinant attenuated bacterial vaccines must be fully attenuated to prevent disease symptoms and highly immunogenic. Conventional in-frame gene deletion can over attenuate the strain, making it more susceptible to host immune defenses and limit its ability to colonize internal tissue (4). To overcome these problems, our lab has previously demonstrated the ability to effectively attenuate E. piscicida through the use of regulated-delayed attenuation systems (5,6). The promoters of the fur and crp genes were replaced by a tightly-regulated araC P araBAD cassette such that expression was arabinose dependent during growth in vitro. This feature enables the strain to phenotypically mimic the virulent of wild-type E. piscicida strain at the time of vaccination and enhance the tissue colonization. Inside host tissues, absence of arabinose prevents the synthesis of the Fur and Crp proteins, allowing attenuation to gradually manifest and prevent disease symptoms (4). The safety and control of these strains can be further enhanced by replacing the promoters of several virulence factors with externallyregulated promoters. Additionally, these recombinant attenuated Edwardsiella vaccines (RAEVs) are capable of stimulating both innate and adaptive immune responses in fish (5,6). Zebrafish vaccinated with an RAEV showed up-regulation in the transcription of genes for several pro-inflammatory cytokines, including TNF-a and IL-1b (5,7). Additionally, immunized fish had significantly higher levels of serum IgM and increased levels of protection against challenge with the wild-type strain (5). These attributes suggest that these RAEV strains would be optimal delivery systems for heterologous recombinant proteins.
The ability of an organism to act as a recombinant protein producer is contingent on its ability to effectively retain the expression vector. To ensure that the vector is not lost, a selective pressure is created so that only organisms harboring the vector with the selective marker are able to grow. In laboratory settings, antibiotics typically serve as the selective agent. However, vectors with antibiotic selection markers may not be retained in vivo when the selective pressure is removed. Thus, vectors used in live-attenuated vaccines should have a selective marker that is applicable in host tissue. Diaminopimelic acid (DAP) is a necessary precursor for peptidoglycan synthesis in Gramnegative bacteria (8). Aspartate ß-semialdehyde dehydrogenase is an enzyme required for the production of DAP and encoded for by the asdA gene (9,10). Gram-negative bacteria lacking asdA are unable to synthesize DAP and undergo lysis if they cannot obtain it from their surroundings (11)(12)(13)(14). DAP is not synthesized nor metabolized in mammalian tissue and the nutritional requirements of fish suggests they also lack biological pathways that utilize DAP as a substrate (15,16). This makes asdA an appropriate selection marker for live vaccines and is the basis for the balanced-lethal system, in which asdA deficient bacteria are dependent on a vector containing the cloned wild-type asdA gene for survival (17). This system is utilized in many Recombinant Attenuated Salmonella Vaccines (RASVs) to deliver protective antigens against multiple pathogens through antigen secretion or surface display (18,19). Additionally, E. ictaluri and E. tarda DasdA mutants are capable of expressing heterologous antigens harbored on AsdA + vectors (14,20). These findings suggest that an RAEV possessing an asdA deletion could also function in a same manner.
Ichthyopthirius multifiliis is a ciliated protozoan and the etiological agent of white spot disease in fish. Infection by I. multifiliis has been shown to elicit various primary and secondary immune responses in catfish that promote parasite clearance (21)(22)(23)(24). These adaptive immune responses in fish suggests that vaccination would be a safe and effective method to prevent disease. Surface immobilization antigens (IAGs) expressed by I. multifiliis are immuno-dominant and have become candidate proteins for vaccine development (23). Channel catfish and rainbow trout immunized with DNA vaccines encoding IAG52B produced specific antibodies, but showed no significant increase in the level of protection in challenge studies, suggesting that naked DNA alone is insufficient to induce a strong enough immune response to confer protection (25,26).
The purpose of this study was to construct an antibiotic-sensitive and highly immunogenic Recombinant Attenuated Edwardsiella Vaccine (RAEV) vector system with regulated-delayed attenuation in vivo attributes. The asdA gene was deleted and an optimal balanced-lethal system was developed by using diverse recombinant AsdA + plasmids containing different promoter sequences and origins of replication to generate as a platform for the heterologous antigen delivery system. The I. multifiliis gene encoding the surface immobilization antigen 52B (IAG52B) was codon optimized to enable high-level expression in Edwardsiella. In addition, a codon specifying incorporation of glutamine in Edwardsiella was specified to replace the nonsense codon used to specify glutamine in I. multifiliis. This modified sequence was then inserted into an AsdA + expression vector and fused with the blactamase type II signal sequence for better secretion of the cloned gene products into the periplasm. The strain was then evaluated for its ability to elicit innate and adaptive immune responses towards both E. piscicida and I. multifiliis in immunized zebrafish.

Bacterial Strains, Plasmids, Media, and Reagents
All bacterial strains and plasmids used in this study are listed in Table 1. Bacterial strains were grown on Luria-Bertani (LB) agar or in LB broth. Where necessary, media was supplemented with 15% agar, 10% sucrose, colistin sulfate (Col) (12.5 µg/ml), chloramphenicol (Cm) (25 µg/ml), diaminopimelic acid (DAP) 50 µg/ml. Growth of bacteria was determined by spectrophotometrically and/or by plating following serial dilution. Oligonucleotides were from IDT (Coralville, IA). New England BioLabs restriction endonucleases and T4 ligase were used for cloning. For all PCR reactions, GoTaq DNA polymerase (Promega, catalog# M3008) was used. For plasmid DNA isolation and purification of gel fragments and PCR products, Qiagen products (Hilden, Germany) were used.

Construction of asdA Mutants
To develop antibiotic-sensitive strains of live-attenuated recombinant bacterial vaccines, a balanced-lethal host-vector system was constructed by deletion of the aspartate bsemialdehyde dehydrogenase (asdA) gene. The DasdA10 defined deletion mutation encompasses a 1100 base pair deletion including the ATG start codon but not does not include the last four bases, "CTAG", specifying the stop codon for the gene. Primers asdA-1F-XbaI and asdA-2R ( Table 2) were designed to amplify the upstream of the asdA gene flanking region (432 bp). The downstream asdA gene flanking region (583 bp) was amplified by primers asdA-3F and asdA-4R-KpnI ( Table 2). A XbaI site was included in asdA-1F-XbaI, and a KpnI site was included in asdA-4R-KpnI. The flanking regions were amplified from E. piscicida J118 genomic DNA. The two PCR fragments were joined by overlapping PCR with primers asdA-1F-XbaI and asdA-4R-KpnI, and the products were cloned into the XbaI/KpnI site of the suicide vector pRE112 (29). The resulting plasmid was designated pG8R8000. To construct the E. piscicida DasdA10 mutant, the suicide plasmid was transferred from Escherichia coli c7213 to E. piscicida wild-type strain J118 through conjugation. LB agar plates containing Col, Cm, and DAP were used to isolate strains with single-crossover plasmid insertions. A sacB-based sucrose sensitivity counter-selection system was used to select for bacteria that had lost the suicide vector after a second homologous recombination (i.e., allelic exchange) (29). The colonies were screened for growth in the presence of DAP, as well as for Cm S , and Col r . The DasdAdeletion mutant was confirmed by PCR and DNA sequencing.

Construction of RAEV Strains With Regulated Delayed Attenuation Phenotype
To construct the E. piscicida DasdA10 strain (c16000) with a regulated-delayed attenuation phenotype, DP fur and DP crp deletion insertion mutations were added sequentially to c16000 by using suicide plasmids pG8R8024 and pG8R8009 as described previously (5,6). The resultant strain had a genotype DasdA10 DP fur170 ::TT araC P araBAD fur DP crp68 ::TT araC P araBAD crp and numbered as c16022.

Construction of AsdA + Plasmids for Complementation of E. piscicida DasdA Mutant Strains
A series of different AsdA + plasmid vectors were constructed with pUC ori and pBR ori containing the E. piscicida asdA gene with modifications of the asdA promoter, SD sequence and start codon. The E. piscicida asdA gene was amplified with its wild-type promoter and Shine-Dalgarno (SD) sequence, or with only the SD asdA sequence and also with modification of the start codon from ATG to GTG or the asdA gene without its SD sequence by using primers listed in Table 2. Forward and reverse primers were tagged with restriction enzyme sites for XbaI and KpnI. Fragments of the pYA3341 (pUC ori) (30) and pYA3342 (pBR ori) plasmids (30) minus the S. Typhimurium asdA gene were amplified by PCR with the primer pair P42F-KpnI and p42R-XbaI ( Table 2). After gel purification, fragments were ligated with T4 DNA ligase and transformed into the E. piscicida DasdA strain c16000 and plated on LB agar plates. The recombinant plasmids were confirmed by restriction digestion with XbaI and KpnI and sequencing. The resulting plasmids were named pG8R8011, pG8R8012, pG8R8013, pG8R8014, pG8R8015, pG8R8016, pG8R8017 and pG8R8018 ( Table 1).

Growth Curve Analysis
The growth of the E. piscicida DasdA mutant strain harboring different AsdA + vectors was analyzed and compared to the mutant strain without a plasmid in LB both in presence and absence of DAP. A series of plasmids were constructed with differing copy numbers, presence or absence of the promoter for the asdA gene and with either ATG or GTG start codons as described in Table 1. Standing overnight 30°C cultures (OD 6000 .6) of E. piscicida strains were diluted 1:100 into prewarmed LB or LB plus DAP broth and incubated at 30°C with shaking at 180

Primer
Sequence Restriction enzyme sites were underlined. Start codon and SD sequences were in bold letter.
RPM. The OD 600 was measured every 60 min. The growth curves were calculated using the automated growth curve device Bioscreen C (Growth Curves USA, Piscataway, NJ).

Determination of Lethal Dose 50 (LD 50 )
To determine the LD 50 of E. piscicida strains i.e. J118, c16000, c16010, c16012 and c16012(pYA3493) ten-fold serial dilutions of E. piscicida fresh cultures were made in sterile BSG, and the concentration of bacteria was determined by the spread-plate method. Fish were i.c. injected in a dose of 10 µl of BSG containing different concentrations of CFU/fish ( Table 3) with 10 fish in each group (two replicate tanks, 5 fish in each tank). Mortality was documented daily over a 15-day period, and the Reed and Muench method was used to calculate the LD 50 values (34).

Colonization of RAEV Strain in Zebrafish
Zebrafish were inoculated with~5 x 10 3 cells of c16022, carrying the Ich IAG52B antigen gene, in a dose of 10 µl of BSG by i.c.
injection. The spleen and kidneys were collected at days 3 and 5 after inoculation. Tissues were homogenized in 200 µL of BSG. A 10-fold serial dilution of each sample was used and plated on LB agar plates containing 0.2% arabinose and 10 µg/mL of colistin sulfate. The plates were incubated at 30˚C for 48 h and then the colonies were counted to determine the bacterial load in each organ. The results obtained from five fish (n = 5) in each time point were plotted with the bars indicating standard error.

Immune Protection Against Wild-Type E. piscicida Challenge
The immune protection mediated by c16022 against wild-type E. piscicida challenge was evaluated following i.c. vaccination, with a dose of 1 x 10 4 cells/fish using 15 fish/group. Fish were given a booster dose of 1 x 10 4 cells/fish after 2 weeks. Control fish were injected with 10 µL of BSG. Two weeks after receiving the booster dose, fish were challenged by i.c injection of virulent, wild-type E. piscicida J118 at a dose of 1 x 10 5 cells/fish. Mortalities were recorded for 14 days and represented as percent survival. The experiment was repeated once using 15 fish/group. Since the

Subcellular Fractionation
The periplasmic fraction was prepared by following the method described by Kang et al., 2002 (30). E. piscicida and E. coli cells harboring the pG8R8029 plasmid were grown in LB broth at 30˚C or 37˚C up to OD 600 of 0.8 and centrifuged at 6000 g for 10 min. The supernatant fluid was saved for analysis of secreted proteins. Equal volumes of periplasmic, cytoplasmic, outer membrane, supernatant fractions and total lysate samples were separated by SDS-PAGE for western blot analysis.

SDS-PAGE and Western Blotting
Plasmid encoding IAG52B (pG8R8029) and control plasmid (pYA3493) were electroporated into E. piscicida or E. ictaluri or E. coli cells. Strains were grown in LB broth and 0.2% arabinose was added when necessary. Bacterial cells were grown at 30°C or 37°C with aeration (180 rpm) to an optical density at 600 nm (OD 600 ) of 0.8. For the western blot analysis, 1 ml of bacterial culture was centrifuged, suspended in 100 µl of phosphate buffered saline (PBS; pH 7.4) and mixed with 100 µl of 2X SDS loading buffer. Protein samples were boiled for 10 minutes, and then 10 µl samples were loaded onto a 12% SDSpolyacrylamide gel electrophoresis (SDS-PAGE) gels and electrophoresed. Samples were transferred onto nitrocellulose membranes. Membranes were blocked overnight at 4°C using fat-free milk powder dissolved in phosphate buffered saline (PBS) (5%, wt/vol) supplemented with 0.05% Tween 20 (PBS-T). The membranes were incubated with a primary rabbit polyclonal anti-Ich antibody (35). Membranes were washed with PBS-T three times, and then incubated with an alkaline phosphatase-conjugated anti-rabbit immunoglobulin G (IgG) (Sigma) diluted 1:10,000 in blocking buffer at room temperature for 1 h. A mixture of nitroblue tetrazolium and 5bromo-4-chloro-3-indolylphosphate (NBT-BCIP) (Amaresco), chromo-genic substrates for alkaline phosphatase was used to develop color. The reaction was stopped after 10 min by washing with several large volumes of deionized water.

RNA Isolation and First-Strand cDNA Synthesis
Total RNA was extracted from zebrafish tissue samples (i.e., gill, kidney, intestine, and spleen) using TRIzol ® (Ambion) following the standard protocol (Invitrogen). The total RNA concentration was measured by a UV-spectrophotometer (NanoDrop2000c, Thermo), and the relative purity was analyzed by the ratio of the absorbance value at 260 nm (A260) divided by the absorbance value at 280 nm, with a value of~2.0 being considered highly pure. 3 µg of total RNA was then treated with 1 U of Thermo Scientific ™ DNase I, RNase-free (FEREN0521) to remove any residual DNA from the sample. Reverse transcription was carried out using oligo-dT primer and Thermo Scientific ™ RevertAid ™ Premium First Strand cDNA Synthesis Kit (Catalog #FERK1622). cDNA synthesis was confirmed by PCR amplification of the ß-actin gene, keeping DNase-treated RNA as a negative control. Until further analysis, the synthesized cDNA was stored at -80˚C.

Quantitative Real-Time PCR (qRT-PCR) Analysis
To evaluate the ability of c16022 specifying the Ich IAG52B antigen to stimulate an immune response, the expression of the genes encoding IL-1b, IL-6, IL-8, and TNF-a in zebrafish tissues was analyzed by qRT-PCR. A 10 µL reaction mixture was prepared consisting of 5 µL of 2 x PowerUp ™ SYBR ™ Green Master Mix (Thermofisher Catalog # A25742), 3.5 µL of PCR grade H 2 O, 0.25 µL of FW and RV primers (2.5 mM each), and 1.0 µL of cDNA. The reaction was carried out in a Quantstudio 3 thermocycler (Applied Biosystems) in three separate wells with conditions of initial denaturation at 95°C for 10 min followed by 45 cycles with denaturation at 94°C for 10 s, annealing at 58°C for 10 s and extension at 72°C for 10 s. The b-actin gene was used for the internal normalization and reaction mixtures without the cDNA template served as the negative controls. PCR efficiencies were determined by analyzing cDNA serial dilutions. Since the efficiencies were almost 100%, the 2 −DDCT method (36) could be used to calculate the relative gene expression compared to the Bactin gene reference. The relative expression ratios were obtained by normalizing expression of the target gene, as determined by mean crossing point (cp) deviation by that of B-actin. The single band amplification and correct size was confirmed by ethidium bromide-stained 1% agarose gels using 8 µL of the qRT-PCR product. The qRT-PCR data was expressed as a mean of three individual experiments with standard error. To determine the significant difference between control and treated groups, Student's t-test was conducted using Microsoft Excel 2010 with a significance level of P < 0.05.

Determination of IgM by ELISA
ELISA was used to assay antibodies in gill, skin and serum to E. piscicida LPS and Ich membrane protein. Samples were prepared as described previously (5)

Sequence Analysis, Phylogenetic Tree and 3D Model of AsdA
The E. piscicida EIB202 (J118) asdA open reading frame consisted of 1104 base pairs (bp) that encoded a putative 368 amino acid (aa) residue protein with an estimated molecular mass of 40 kilodaltons (kDa). To explore the evolutionary development history of bacterial asdA genes, a phylogenetic tree was constructed based on the amino acid sequences ( Figure 1A). The 3D structure of the E. piscicida, E. ictaluri and S. Typhimurium AsdA were predicted by using the Phyre2 web portal. Our results revealed that they share high sequence homology and with similar 3D structures. Each protein showed 16 a-helices and 9 b-sheets that were connected to each other by loops ( Figure 1B).

Construction and Characterization of E. piscicida asdA Mutant
To develop antibiotic-sensitive strains of live-attenuated recombinant bacterial vaccines, we used a balanced-lethal hostvector system to delete the aspartate b-semialdehyde dehydrogenase (asdA) gene. The DasdA10 defined deletion mutation encompasses a 1100 base pair deletion including the ATG start codon but does not include the last four bases, "CTAG", of the gene. The upper panel of Figure 2A illustrates the chromosomal structures of the wild-type and mutant strains. Construction of the E. piscicida DasdA strain was done through suicide plasmid mediated homologous recombination, using suicide vector pG8R8000, a pR112 (Cm) based suicide vector ( Table 1), by allelic replacement in the parent strain J118. The genotype was confirmed via PCR ( Figure 1B) and sequencing. Agarose gel analysis and sequencing of PCR product confirmed that DasdA was 1100 bp smaller than the wild-type strain. The resultant E. piscicida DasdA mutant was named c16000 ( Table 1). The phenotype was verified by growth of DasdA mutant (c16000;) in presence or absence of DAP ( Figures 2C,  D). The DasdA mutant could not survive in LB agar plates without being supplemented with DAP, but in the presence of DAP, the DasdA mutant strain grew like the wild-type strain. These results confirm that the E. piscicida asdA gene is functional and might work in a similar manner to that observed in E. ictaluri (14) and Salmonella (17).

Construction of E. piscicida Vaccine Delivery Strains With Regulated-Delayed Attenuation Phenotype
Deletion of either fur or crp attenuated E. piscicida (5, 6) and E. ictaluri (37,38). Strains were constructed with deletion insertion mutations that conferred a phenotype of regulated-delayed attenuation in vivo (5). The principle of regulated-delayed attenuation in vivo is based on the replacement of the fur and crp gene promoters with a tightly regulated araC P araBAD cassette such that the expression of these genes is dependent on arabinose supplied during growth. Thus, following internal tissue colonization, Fur and Crp protein synthesis is terminated due to the absence of arabinose and attenuation is steadily manifested in vivo to prevent disease symptoms. The promoters, including all sequences that interact with activator or repressor proteins, for the fur and crp genes were deleted, and the improved araC P araBAD cassette was substituted in E. piscicida DasdA strain (c16000) to yield a strain with genotype DasdA10; DP fur170 ::TT araC P araBAD fur; DP crp68 ::TT araC P araBAD crp (P stands for promoter and TT for transcription terminator), and the strain was designated c16022.

Construction AsdA + Vectors to Develop a Balanced-Lethal System in E. piscicida
Gram-negative bacteria with asdA mutants have an obligate requirement for diaminopimelic acid (DAP), which is a crucial component of the peptidoglycan layer (4). In surroundings deprived of DAP, i.e., animal tissues, they will undergo lysis. Deletion of the asdA gene has previously been used to develop antibiotic-sensitive strains of live-attenuated recombinant bacterial vaccines (14,19). Introduction of an AsdA + plasmid into a DasdA mutant makes the bacterial strain plasmid dependent (17). This dependence on the AsdA + plasmid vector creates a balanced-lethal complementation between the bacterial strain and the recombinant plasmid.
Eight different AsdA + plasmid vectors were constructed with pUC ori or pBR ori containing the E. piscicida asdA gene with modifications of the asdA promoter, SD sequence and start codon ( Figures 3A, B). Plasmids pG8R8011 to pG8R8014 contain the pUC ori whereas plasmids pG8R8015 to pG8R8018 have the pBR ori. Plasmid pG8R8011 and pG8R8015 contain the wild-type asdA gene with the wild-type promoter, which includes the 118 bp upstream sequence from the start codon (ATG). Plasmid pG8R8012 and pG8R8016 contain E. piscicida asdA ORF with Shine-Delgrano (SD) sequence "AGGA". pG8R8013 and pG8R8017 contain only the wild-type asdA ORF sequence. Plasmid pG8R8014 and pG8R8018 encode asdA sequence with the start codon "GTG" and (SD) sequence "AGGA". Each of these plasmids were individually incorporated into the DadsA strain c16000, and growth curves determined ( Figure 3C). All of these E. piscicida AsdA + plasmids were able to complement the asdA gene of c16000. The strain complemented with pG8R8018, which has the pBR ori and encodes the asdA sequence with the start codon "GTG" and (SD) sequence "AGGA" grew similar to the wild-type. c16000 complemented with pG8R8014 grew slightly slower compared to other complemented strains. pG8R8014 and pG8R8018 have similar features other than the origin of replication.

Complementation of E. piscicida D asdA Strains With Salmonella and E. ictaluri AsdA + Plasmids
The E. piscicida AsdA enzyme shares 97% and 81% sequence identity with E. ictaluri and Salmonella, respectivelly. Therefore, compared the ability of the asdA gene from E. ictaluri and Salmonella to complement the E. piscicida DasdA mutant (Figure 4). c16000 complemented with pEZ142 showed a similar growth curve to wild-type J118. A high (pUC ori), medium (pBR & p15A ori) and low-copy (pSC101 ori) number Salmonella AsdA + plasmids were used ( Table 1) to complement E. piscicida DasdA mutants. All of these plasmids efficiently complemented the DasdA chromosomal mutation; however, the growth of these strains was slower than those complemented by plasmids with the E. piscicida AsdA+ encoding sequences. A slower growth of c16000(pYA3337) was observed compared to other strains. The slower growth might be due to the low-copy number plasmid pYA3337, which may not produce a sufficient amount of AsdA enzyme needed for the growth of the bacteria. IAG52B ORF consisted of 1383 base pairs and encoded a protein of 460 aa. Different domains of IAG52B were predicted by the InterPro web portal (https://www.ebi.ac.uk/interpro/). IAG52B contains a C-terminal signal IP of 20 amino acids, a middle non-cytoplasmic domain of 422 amino acids and a Nterminal transmembrane region of 18 amino acids. Transgene expression of i-antigen genes imposes a challenge in heterologous systems such as bacteria, because Ich, like other hymenostome ciliates, uses a nonstandard genetic code (39)(40)(41) in which the stop codons UAA and UGA encode glutamine. This problem was addressed by synthesizing and inserting i-antigen gene constructs with altered codons that were optimized for expression in E. coli and E. piscicida ( Figure 5A). The composition of the IAG52B amino acid is depicted in Figure 5B. The secondary structure of IAG52B ( Figure 5C) was predicted by Protean software, and the 3D structure ( Figure 5D) was predicted by Phyre2 web portal. Our results indicate that the IAG52B protein contains 20 b-sheets and 7 a-helices and is connected by loops.
The codon-optimized sequence of IAG52B non-cytoplasmic domain of 1269 bp was PCR amplified and cloned into the pYA3493 ( Figure 6A) at the EcoRI-BamHI site. The IAG52B gene was fused into the same reading frame of bla SS under the P trc promoter. The resultant plasmid was named as pG8R8029 ( Figure 6B). Both pYA3493 and pG8R8029 were introduced into the DasdA strains: E. piscicida (c16000), E. ictaluri (J111) and E. coli (c6212). The synthesis of IAG52B was analyzed by western blotting. As shown in Figure 6C, synthesis of IAG52B was detected in E. piscicida and E. coli, but no signal was noticed in E. ictaluri and cells harboring the control plasmid pYA3493.
E. piscicida strain c16000 and E. coli strain c6212 harboring pG8R8029 were analyzed for subcellular localization of IAG52B.
As expected, IAG52B was detectable in cytoplasm, periplasm and the outer membrane in both c16000(pG8R8029) and c6212 (pG8R8029) ( Figure 6D). A high amount of protein was detected in the cytoplasmic fraction but only half of the IAG52B was secreted into the periplasmic fraction in both E. piscicida and E. coli. Interestingly, some of i-antigen associated with the outer membrane in E. piscicida was detected; this suggests that some antigen may be incorporated into outer membrane vesicle (OMVs) that are very immunogenic and excellent means for antigen delivery. Next, the synthesis of IAG52B in the vaccine delivery strain c16022 was analyzed. Large amounts of IAG52B synthesis were detected in c16022 (pG8R8029) and no signal was observed in the vector control c16022(pYA3493) ( Figure 6E). This result confirms that IAG52B is synthesized in the E. piscicida vaccine strain c16022 (pG8R8029). A similar subcellular localization pattern was observed in c16000(pG8R8029).
To examine the stability of plasmids pYA3493 and pG8R8029 in E. piscicida c16022, strains c16022(pYA3493) and c16022 (pG8R8029) were cultured with daily passage of 1:1,000 dilutions for five consecutive days in LB broth containing DAP and arabinose. Hundred colonies were screened daily for DAP sensitivity in LB agar plates in the presence or absence of DAP. All the colonies were DAP independent, indicating that pYA3493 and pG8R8029 are very stable in the c16022 vaccine strain. IAG52B synthesis was analyzed by western blotting from cells obtained from the last day of culture of the stability test, and there was a detectable signal, which suggests both pG8R8029 in the vaccine strain and antigen expression are stable.

Verification of Regulated-Delayed Attenuation of E. piscicida Mutant Strains in Zebrafish
The virulence of the E. piscicida vaccine and wild-type strains was determined by LD 50 values in zebrafish injected intracoelomically with different concentrations of bacteria. Our result showed that the vaccine strain c16022 harboring the The secondary structures of IAG52B protein were predicted by Protean software by using Gramier-Robson and Chou-Fasman methods. Lines 1, 3, 5, and 7 are the Gramier -Robson methods; the red represents the alpha helix, the green represents the beta fold, the blue represents the turn, the yellow represents the random coil. Lines 2, 4, 6 are for the Chou -Fasman method; the red represents the alpha helix, the green represents the beta fold, the blue for the turn, without random coil prediction. (D) The 3D structure of IAG52B protein was predicted using the Phyre2 web portal.
control plasmid pYA3493 had significantly higher LD 50 levels compare to the wild-type strain ( Table 3). The LD 50 of the wildtype strain J118 was 1.1 × 10 4 CFU, while the LD 50 of the c16022 (pYA3493) increased up to 2×10 5 CFU, which is more than a 10fold increase ( Table 3). As expected, the LD 50 of the c16022 (pYA3429) was similar to the LD 50 of c16022(pYA3493) (data not shown). This result indicates that the E. piscicida crp gene product contributes to the virulence and pathogenicity of E. piscicida. The LD 50 of c16012 was 2 x 10 4 CFU, which was similar to the wild-type strain J118. This result confirms that DP fur170 ::TT araC P araBAD fur deletion insertion mutation may not affect the virulence of E. piscicida by i.c. injection.
As expected, the LD 50 of the c16022(pYA3429) was similar to the LD 50 of c16022(pYA3493) (data not shown).

Dissemination and Colonization of c16022(pG8R8029) in Zebrafish Tissues
To examine the ability of c16022(pG8R8029) to colonize and disseminate into fish tissues, zebrafish were immunized with the vaccine strain c16022(pG8R8029) by i.c. injection. The recovery of the vaccine strain from the kidney and spleen was determined at days 3 or 5 post-vaccination. A significant number of c16022 (pG8R8029) was recovered in both the kidney and spleen at days 3 or 5 of post-vaccination. However, the number of bacteria was higher at day 3 compared to day 5 ( Figure 7). This result indicates that the vaccine strain successfully colonized and disseminated into different tissues of fish.

Assessment of Cytokine Gene Expression in Different Organs of Immunized Zebrafish
To investigate the efficiency of the vaccine strain in modulating immune responses of zebrafish, the expression of genes associated with the immune response including tumor necrosis factor-a (tnf-a), interleukin 1b (il-1b), interleukin-6 (il-6) and interleukin 8 (il-8) in the gills, kidney, intestine and spleen were assessed by qRT-PCR. All of these cytokine genes were significantly upregulated in the tissues of vaccinated fish compared to unvaccinated fish. Furthermore, their expression was higher at day 3 compared to day 5 and 7 in all tested tissues of the vaccinated fish. A significant increase in tnf-a gene expression was detected in kidney and spleen ( Figure 8A). il-1b gene expression levels were significantly increased in all organs after 3 days post-vaccination. No changes were found in the expression levels of il-1b at day 5 and 7 in vaccinated fish compared to the control group ( Figure 8B). The expression of il-6 and il-8 patterns were similar to il-1b, showing significant upregulation in all tissues at day 3, whereas, the changes at day 5 and day 7 were insignificant compared to control tissues ( Figures 8C, D).

Antibody Responses to E. piscicida LPS and Ich Membrane Protein
Serum and mucosal immunoglobulin M (IgM) responses to E. piscicida lipopolysaccharide (LPS) and Ich membrane protein were measured by ELISA at 4 weeks (before challenge) and 6 weeks (2 weeks after challenge) post-vaccination of c16022 (pG8R8029). The LPS and Ich specific IgM titers were induced in both serum and mucosal tissues of vaccinated fish and compared to unvaccinated control fish. There was a significant upregulation of anti-LPS IgM titers in the serum at both 4-and 6-weeks post-vaccination compared to control group. The anti-Ich IgM titer was significantly increased only at 6 weeks of postvaccination ( Figure 9).

DISCUSSION
The advantage of using live bacterial vaccine vectors is their capability to invade and colonize internal tissues after mucosal delivery, which is critically important for inducing immune protection (32). Bacterial vaccine vectors have been used to produce protection against self as well as heterologous antigens (42)(43)(44). Our lab has developed a strategy of regulated-delayed attenuation in Salmonella, which makes the vaccine vectors safe and immunogenic for delivery of heterologous antigens to mice (45)(46)(47). We recently established the regulated-delayed attenuation phenotype in E. piscicida by using the arabinose- regulated promoter (5,6). In this study, the promoters of both fur and crp genes were replaced with a tightly regulated araC P araBAD cassette such that expression of these genes is dependent on arabinose which is provided during growth in vitro. Fur and Crp proteins cease to be synthesized due to the absence of arabinose and attenuation is progressively established in vivo to preclude induction of disease symptoms. Our study indicated that the E. piscicida strain carrying these two deletion insertion mutations is attenuated in zebrafish by i.c. injection. To eliminate the need for an antibiotic-resistance selection marker for maintenance of plasmids in bacterial cells, our lab developed the balanced-lethal vector-host systems in Salmonella and E. ictaluri to ensure maintenance of the plasmid vectors in vivo and in vitro, without reliance on antibiotic-resistance markers (12,14,17). To develop a balanced-lethal system in E. piscicida, the asdA gene was characterized, whose product is essential for the synthesis of DAP, a crucial element of bacterial cell wall peptidoglycan. E. piscicida DasdA strain c16000 can only survive when DAP is supplied in the growth media or contains a plasmid with the wild-type asdA gene. Since DAP is not present in in vivo environments, the DasdA strain will undergo lysis unless a plasmid with the asdA gene is present within the bacteria, establishing a powerful selection marker for maintenance of the plasmids within the RAEV, as observed previously in Salmonella and E. ictaluri (12,14,17). The ideal balanced-lethal system should have similar level of virulence as the wild-type strain. The amount of AsdA synthesis by the Asd + vector in RASV affects the growth and immunogenicity of the strain (30). E. piscicida AsdA shares a high percentage of sequence identity and structural identity with, E. ictaluri (97.28%) and Salmonella (82.83%). As expected, plasmids containing the wild-type asdA gene from E. ictaluri and Salmonella were able to complement and establish the balanced-lethal system in E. piscicida DasdA strains. The pYA3493 AsdA + vector system has been successfully used in live recombinant Salmonella and E. ictaluri vaccines (14,30,31,48). Therefore, the pYA3493 AsdA + vector system was used to deliver the antigens by RAEV. Vaccine efficacy depends on successful delivery of protein antigens for presentation in an optimal form and it has been observed that secreted proteins are immunogenic and interact with antigen-presenting cells because of their subcellular locations (49). b-Lactamase is a well-A B C D FIGURE 8 | tnf-a, il-1b, il-6 and il-8 gene expression in c16022(pG8R8029) vaccinated and control zebrafish. Zebrafish were vaccinated with c16022(pG8R8029) by i.c. injection. At three, five and seven days of post-immunization, total RNA was extracted from the gills, kidney, intestine and spleen and cDNA was prepared. The qRT-PCR assay was conducted to analyze the expression of tnf-a, il-1b, il-6 and il-8 genes using b-actin as an internal control. The results are expressed as mean ± standard error (bars) from three separate experiments. Differences between uninfected (control) and infected groups were analyzed by two-way ANOVA, where asterisks (*) indicate significant difference (*P < 0.05, **P < 0.01, ****P < 0.0001) with respect to the control group.
characterized periplasmic secreted protein in Gram-negative bacteria and translocation depends on b-Lactamase signal sequence (bla SS) composed of 23 amino acid (aa) N-terminal residues (50,51). The pYA3493 plasmid contains the P trc promoter region, b-lactamase signal sequence (bla SS), and a multicloning sites. It was designed for secretion of recombinant antigens to the periplasm from the cytoplasm (48). Infection by I. multifiliis has been shown to elicit both primary and secondary immune responses in catfish that promote parasite clearance (21)(22)(23)(24). This adaptive immune response in fish suggests that vaccination would be a safe and effective method to prevent white spot disease. Surface immobilization antigens (IAGs) expressed by I. multifiliis are immuno-dominant and have become candidate proteins for vaccine development (23). Channel catfish and rainbow trout immunized with DNA vaccines encoding IAG52B produced specific antibodies, but showed no significant increase in the level of protection in challenge studies, suggesting that naked DNA alone is insufficient to induce a strong enough immune response to confer protection (25,26). We expect that the Ich antigen delivered by RAEV should solve this problem. Sera from immunized fish immobilize I. multifiliis theronts in vitro demonstrating a role for i-antigens in protective immunity (52). Cytokines play an important role in immune surveillance against and Ich IAG52B membrane protein were measured by ELISA at 4 weeks (4w) post-vaccination with c16022(pG8R8029) and 2 weeks after J118 challenge (6w). The results are expressed as mean ± standard error (bars) from three separate experiments. Differences between treated and control groups were analyzed by two-way ANOVA, where asterisks (*) indicate significant difference (**P < 0.01, ***P < 0.001). bacterial infection in fish (53)(54)(55). It was observed that, RAEV-Ich strain induced cytokine tnf-a, il-1b, il-6 and il-8 gene expression in different tissues of zebrafish. Activation of proinflammatory cytokines is involved in regulating immunoglobulin synthesis in teleosts (5,7,56). Zebrafish receiving booster doses of a live attenuated Vibrio anguillarum vaccine at 2 weeks after primary vaccination were better protected against vibriosis in comparison to fish that received a single vaccination (57). In the current study, we delivered booster immunizations at 2 weeks after primary vaccination to ensure long-term protective immunity. We did not detect signs of disease in vaccinated fish during the immunization periods. Our study demonstrates that RAEV-Ich induced a mucosal and serum anti-IAG52B IgM titer in zebrafish. As expected, RAEV-Ich vaccine strain also induced E. piscicida protective immunity in the immunized zebrafish. Immunized fish showed elevated levels of E. piscicida anti-LPS IgM in both mucus and serum. The immune protection mediated by RAEV-Ich vaccine strain c16022(pG8R8029) and vector-control strain c16022(pYA3493) against E. piscicida infection were evaluated. The survival rate was higher than 60% and 55% in zebrafish immunized with c16022(pG8R8029) and c16022(pYA3493) respectively. This confirms that RAEV confers immune protection against the virulent E. piscicida.
In summary, we have invented and developed an innovative antibiotic-sensitive Recombinant Attenuated Edwardsiella Vaccine (RAEV) vector system with in vivo display of regulated-delayed attenuation. The bla SS-IAG52B fusion protein was synthesized and secreted into the RAEV periplasm and delivered it to fish. RAEV-Ich induced different levels of immune responses and gave significant protection to the fish against E. piscicida infection. RAEV-Ich also induces anti-IAG52B IgM in zebrafish suggesting this vaccine strain may protect fish against white-spot diseases. This multidisciplinary approach using cutting-edge technologies will address the sustainability challenges of aquaculture increased food security.

ETHICS STATEMENT
The animal study was reviewed and approved by University of Florida IACUC.